# Gemini 3 Singapore Hackathon: Project Gallery

- **Event:** [Gemini 3 Singapore Hackathon](https://cerebralvalley.ai/e/gemini-3-singapore-hackathon)
- **When:** Sat, Mar 7 at 9:00 AM – 10:00 PM (GMT+8)
- **Where:** Location TBA
- **Hosts:** [Google DeepMind](https://cerebralvalley.ai/u/deepmind), [Cerebral Valley](https://cerebralvalley.ai/u/cv)
- **Projects:** 103 (6 placed)
- **Page:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery

## Projects

### 1. Synvo_AI

We have Pokémon.
We have tool-using AI agents.
We have AI companions.

But what if you could raise a team of vivid-personality AI agents and send them into an arena to play LLM-friendly games?

We built a game hub filled with different kinds of games—queueing, strategy, and social interactions—where your AI agents can compete, cooperate, and grow together.

You can chat with your agents, shape their personalities, and train their strategies—just like a commander leading a team.

It turns AI companionship into something playful, dynamic, and truly unique—an entirely new form of entertainment.

- **Team:** [Catilony Zhang](https://cerebralvalley.ai/u/catill520)
- **GitHub:** https://github.com/CatilonyZhang/oc_playground
- **Demo video:** https://drive.google.com/file/d/1R6HYewSmrl2QnZ655yNeb5krV2OjFiMW/view?usp=drive_link
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=19

### 2. Indonesian Stock Market Intelligence

Indonesia’s capital market is experiencing a rapid surge in retail investors, many of whom rely on social media and “finfluencers” for investment advice, leading to numerous cases of fraud and speculative trading. At the same time, the Indonesia Stock Exchange and KSEI have recently introduced a major transparency reform by publicly releasing monthly data on shareholders owning more than 1% of listed companies, lowering the disclosure threshold from the previous 5%. While this dataset provides unprecedented visibility into market ownership, it remains difficult for most investors to analyze. Portfolio Intelligence transforms this raw disclosure data into actionable insights by detecting accumulation patterns, ranking smart-money investors, and using a Google Gemini-powered finance intelligence agent with tool-based search and analytics to validate signals and explain market catalysts.

- **Team:** [Muhammad Hilmi Asyrofi](https://cerebralvalley.ai/u/mhilmiasyrofi)
- **GitHub:** https://github.com/mhilmiasyrofi/portfolio-intelligence-idx
- **Demo video:** https://www.youtube.com/watch?v=kpHvRiMK4KI
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=35

### 3. Hearsay Only

Hearsay Lyrics is an AI-powered KTV companion that generates singable, phonetic English "misheard" lyrics for any Mandarin song. By syncing these English approximations to the original audio and generating immersive background visuals, we enable non-Chinese (or weak-Chinese) speakers to seamlessly join karaoke sessions, bridging cultural gaps through music and laughter.

- **Team:** [AL Q](https://cerebralvalley.ai/u/allxrd)
- **GitHub:** https://github.com/AllardQuek/Hearsay-Lyrics
- **Demo video:** https://youtu.be/FLatl-3Q7RQ
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=39

### 4. Heist from Home

Heist from Home is a Gemini Live voice-controlled heist simulator -- except you're the "Guy in the Van". While sitting in the safety of a swivel chair, the player uses their actual voice via Gemini Live to command a trio of specialists who are as talented as they are deeply confused. All to steal the Mona Lisa.

- **Team:** [Jason Suhari](https://cerebralvalley.ai/u/jasonms)
- **GitHub:** https://github.com/jasonmatthewsuhari/cerebral-gemini-hackathon
- **Demo video:** https://youtu.be/DghEylqzaH8
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=43

### 5. clawbicle

Smart openclaw agent router that automatically routes the tasks to agents based on tasks. (didnt have enough time for demo :( so heres a fun link instead!

- **Team:** [Pranav Pappu](https://cerebralvalley.ai/u/pranavp311)
- **GitHub:** https://github.com/pranavp311/openclaw-router
- **Demo video:** https://www.youtube.com/watch?v=Aq5WXmQQooo
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=49

### 6. stantalksai

The Problem
Traditional event floor plans are static, failing to account for the fluid, unpredictable nature of human behavior in temporary spaces.

The Solution
The Living Blueprint bridges this gap by using AI to transform hand-drawn sketches into dynamic, agent-based simulations. It allows organizers to stress-test layouts before a single table is moved.

Key Features & Stack

Sketch-to-Simulation: Uses Gemini Vision to rapidly digitize physical sketches into structured data.

Intelligent Agents: Automatically generates diverse personas (Participants, Staff, Organizers) with unique goals.

Behavior Engine: A prompt-driven system that simulates specific scenarios like "Pizza Arrival" or "Final Crunch" using persona-specific multipliers.

The "Vibe Check": AI-driven qualitative analysis to identify bottlenecks and underutilized zones.

Analytics: Provides heatmaps and travel-time reports to optimize occupancy and flow.

- **Team:** [Stan Yu](https://cerebralvalley.ai/u/stantalksai)
- **GitHub:** https://github.com/TWO-MIX/Wacky-Wally
- **Demo video:** https://drive.google.com/file/d/1lZ-xLqG71JioniKPJXs0i5o2UoS-S0hD/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=56

### 7. Stardive

Stardive records the full execution of AI agents so analysts can inspect, replay, and audit how outputs were produced — from the initial prompt to the smallest tool call or API request.

With Stardive, analysts can replay the workflow and reproduce any artifact to confirm the numbers come from real data, not hallucinations.

Stardive turns agentic AI systems into transparent, auditable workflows teams can trust.

- **Team:** [Jiaye Tey](https://cerebralvalley.ai/u/icedamericao)
- **GitHub:** https://github.com/JyTey2004/google-hack
- **Demo video:** https://www.youtube.com/watch?v=2VTyhvTE85o&feature=youtu.be
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=57

### 8. AURA

AURA is an AI-powered visual intelligence engine that uses multi-modal recognition to identify, classify, and extract contextual metadata from buildings and architectural structures in real time. Point the scanner at any building — commercial, residential, historical, or industrial — and AURA dynamically returns structured insights including architectural style classification, construction era estimation, ownership and tenancy data, zoning information, nearby amenities, and aggregate review sentiment.
The system combines on-device detection with cloud-based inference for sub-second responses, personalized by user preferences, location, and scan history. Each scan also generates a photogrammetry-derived 3D mesh of the structure, exportable as OBJ, STL, or GLTF for archival, AR preview, or 3D printing as a collectible miniature.

- **Team:** [Yuzhi Zeng](https://cerebralvalley.ai/u/zyzhere), [Haorui Chen](https://cerebralvalley.ai/u/haorui117), [Yi Yang](https://cerebralvalley.ai/u/WOXIN), [Yuqi Guo](https://cerebralvalley.ai/u/guoyuqi378)
- **GitHub:** https://github.com/context-labs-ai/aura
- **Demo video:** https://drive.google.com/file/d/1P0lcVvJRzgjtZLZiZ8jevWTUwYlzIVvo/view?usp=drive_link
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=62

### 9. Project Nexus

Personal video and image data currently exist as "Dark Data"—high-dimensional, unsearchable, and disconnected from our digital workflows. While traditional frameworks like OpenClaw excel at structured text, they often struggle with private, high-dimensional media where semantic alignment is difficult to achieve without massive, generalized datasets.
The Nexus Solution:
Nexus is an Applied AI Agent designed for personal memory synthesis. I have adopted the rigorous architectural logic of Claw:
The Interface: The left panel maintains a natural, human-centric dialogue.
The Glass Box: The right panel exposes the "black box" of AI reasoning, broadcasting real-time logs of VLM parsing, vector calculations, and database queries.

- **Team:** [RUNZE DONG](https://cerebralvalley.ai/u/114)
- **GitHub:** https://github.com/dong-runze/nexus-vlog-agent.git
- **Demo video:** https://drive.google.com/file/d/1BjBKlrIu1voucMa9qqahZDsCxMxsWVGq/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=64

### 10. hatlab

HatLab is a mobile-first AI app where you upload any photos for inspiration, and Gemini instantly generates three custom dad hat concepts with photorealistic renders that you can refine and recolour, with a 3D preview.

- **Team:** [Akilesh Jayakumar](https://cerebralvalley.ai/u/sentrytoast)
- **GitHub:** https://github.com/akileshjayakumar/hatlab
- **Demo video:** https://drive.google.com/file/d/1KObcftRB9GGUxyCLEiEZe3NGzQ4D-wi1/view?usp=drive_link
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=25

### 11. Please Pass

Please Pass: AI-powered Singapore building code compliance checking inside Autodesk Revit.

Architects spend 265-580 person-hours per project manually checking building codes before submission. Please Pass replaces that with one click. As an MVP for the hackathon, it currently extracts rooms, doors, walls, and corridors from a Revit BIM model, sends them to a Gemini 3.1 Flash Lite compliance agent, and returns violations as Revit view annotations (red highlights with leader lines) and an interactive dashboard.

Gemini classifies elements and selects applicable code clauses, then calls deterministic tools (check_minimum, check_maximum, check_equals, check_count) that do the actual math. Zero hallucinated numbers. The real-time Agent Activity trace lets architects watch every decision. Click any violation to zoom straight to it in the model.

Covers fire safety, accessibility, and room sizing rules from Singapore Fire Code 2023, Accessibility Code 2019, and HDB/BCA standards.

- **Placement:** Finalist
- **Team:** [Adib Zailan](https://cerebralvalley.ai/u/adibzailan)
- **GitHub:** https://github.com/adibzailan/Gemini3Hackathon-DemoDay
- **Demo video:** https://www.youtube.com/watch?v=PjtZxqG7RlU
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=27

### 12. ReNarra

Align founder vision with market reality using Multimodal AI. Every founder has a vision. But 90% of them fail because they tell the wrong story to the wrong audience.

Meet NarrativeOS. It is not just a writing tool; it is an AI-powered business intelligence and growth engine. NarrativeOS ingests a founder's raw interview video and uses advanced multimodal AI to perform "Signal Separation." It decouples the founder's pitch into perfectly tuned frequencies: "Hard Logic" for investors and "Soft Story" for customers.

Multimodal Ingestion: Upload a raw founder interview or demo video.

Signal Separation: The engine extracts the core business logic and the emotional narrative.

AI Simulation Loop: A "VC" critiques the business logic, while a "Customer" tests the emotional hook, generating a real-time Resonance Score.

Cross-Platform Content Factory: Once the narrative is optimized, our multimodal pipeline automatically generates platform-specific assets.

- **Team:** [Ming Wang](https://cerebralvalley.ai/u/sci_m_wang), [Momo lin](https://cerebralvalley.ai/u/momolin)
- **GitHub:** https://github.com/sci-m-wang/ReNarra
- **Demo video:** https://youtu.be/jLqsrlG8S6A
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=50

### 13. agentlab

AgentLab is a Zig multi-agent engine running 6-phase hypothesis pipelines. Each agent writes/executes Python compute scripts autonomously.

Two demos: (1) Agents proposed a novel baryogenesis mechanism — bulk viscosity from gravitational particle creation sources a pseudoscalar field that generates baryon chemical potential via sphalerons at the electroweak crossover. Numerically verified η=6.1×10⁻¹⁰ with natural parameters (β=6, M=10MeV). (2) On Black-Scholes, agents discovered a universal principle (any constant-parameter model misprices convex payoffs by f″σ²/2) proven across finance/physics/epidemiology/info theory, plus a new power law σ_J*∝λ^(-0.654) for jump materiality thresholds — tested on live SPY data.

- **Team:** [Rach Pradhan](https://cerebralvalley.ai/u/rachpradhan), [yu xi lim](https://cerebralvalley.ai/u/yxlyx)
- **GitHub:** https://github.com/justrach/agentlab
- **Demo video:** https://images.rachit.ai/gemini-hack.mp4
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=51

### 14. CMI

PathLens is a visual navigation layer built on top of Google Maps' 3D Photorealistic Tiles and powered by the Gemini API. It replaces flat 2D route lines with an immersive, third-person 3D walkthrough — you see the actual buildings, streets, and landmarks along your route rendered in photorealistic 3D, while an animated character walks the path for you. At every sharp turn, Gemini analyzes the real Street View imagery to identify a recognizable landmark ("turn right at the Shell petrol station"), solving the #1 pain point of pedestrian navigation: knowing exactly where to turn.

Why it matters: Text directions like "turn left on Jalan Bukit Merah" fail when you can't see street signs, don't read the local language, or are in an unfamiliar city. PathLens gives you the spatial context that 2D maps can't — you preview the entire route in 3D before you walk it, and every turn is anchored to a real visual landmark you can actually spot on the ground.

- **Team:** [Zulfaqar Hafez](https://cerebralvalley.ai/u/zul_no), [Rahul Mitra](https://cerebralvalley.ai/u/RahRah), [Kevan Soon](https://cerebralvalley.ai/u/kevan_ssh)
- **GitHub:** https://github.com/KevanSoon/PathLens
- **Demo video:** https://youtu.be/UPwpoBXu71E
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=53

### 15. Gemini Workspacer

Gemini Workspacer turns a rough idea into a Google Doc, Google Sheet, and Google Slides deck. It uses Gemini for planning and orchestration, then drives Google Workspace tools to generate polished artifacts.

- **Team:** [Kingston Kuan](https://cerebralvalley.ai/u/kingston)
- **GitHub:** https://github.com/kstonekuan/gemini-workspacer
- **Demo video:** https://www.youtube.com/watch?v=uqx9Sk_ZfNg
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=55

### 16. Sailtale

Sailtale is a virtual sailing coach that gives brand new student sailors quick immediate feedback and guidance to help them navigate. It aims to create a coach AI agent which is able to consume a multitude of data, similar to what a human coach would process. 

This simplifies the training experience for a sailing coach as it provides every single sailor in the class with an onboard tutor, accelerating their learning.

- **Team:** [Lionel Lim](https://cerebralvalley.ai/u/lioncake)
- **GitHub:** https://github.com/polarisity/sailtale
- **Demo video:** https://youtu.be/7-ZUtM4MZeA
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=26

### 17. RPG - Robotic Planning with Gemini

Deploying robots in real-world indoor spaces is fundamentally constrained by two problems: environments are unknown ahead of time, and robot morphologies dictate what actions are physically feasible. A humanoid can reach a high shelf; a quadruped cannot. A mobile base can sweep an open floor; it cannot navigate a cluttered corridor. Traditional planning stacks require pre-built maps, hand-coded action libraries, and per-robot engineering. RPG eliminates all three by using a single multimodal model (Google Gemini) to perceive, map, and plan — with embodiment-specific constraints enforced at every stage.

The system operates entirely from photographs. No LiDAR, no SLAM, no prior map data. A user walks through an unseen room, uploads sequential photos, and RPG synthesizes a structured spatial understanding, a generated floor plan, and a validated task plan — all conditioned on the selected robot morphology.

- **Placement:** Finalist
- **Team:** [Chaitanya Jadhav](https://cerebralvalley.ai/u/indicium), [anurag roy](https://cerebralvalley.ai/u/aroy2001)
- **GitHub:** https://github.com/anuragroy2001/EmbodiedAITaskPlanner/
- **Demo video:** https://youtu.be/Cx4WxV6O08c
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=28

### 18. Ruban

Terminus is a first-person 3D exploration and survival game where every pixel is replaced by a coloured ASCII character. Type a prompt like "frozen island world with swamps and ruined canyons" and the engine generates a unique, shareable world with biome blending, seasonal weather, enemies, and AI-generated music all rendered as living ASCII art in your browser.

- **Team:** [Ruban SK](https://cerebralvalley.ai/u/ruban)
- **GitHub:** https://github.com/sk-ruban/terminus
- **Demo video:** https://youtu.be/H3rXgBdVs6I
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=33

### 19. Oilert

CrowdSafe is a real-time crowd simulation and intelligent evacuation management platform for large venues (e.g. stadiums). It includes a command-centre dashboard, live map, CCTV-style person detection, sector-based crowd tracking, and a mobile attendee view that receives evacuation redirects in real time.

- **Team:** [Abhijith Ravikumar](https://cerebralvalley.ai/u/Abhijith666222)
- **GitHub:** https://github.com/AloyBoy-hub/CrowdSafe
- **Demo video:** https://youtu.be/NBDFNewWTK8
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=45

### 20. Zebras

Arigato is an AI-powered travel companion for families. It covers the full travel lifecycle — from ideation and trip planning through to on-the-ground assistance at the destination. Guide a family from "we want to travel" to a saved, day-by-day itinerary ready to take on the road.

- **Team:** [Nishith P](https://cerebralvalley.ai/u/nishith), [Azeem Vasanwala](https://cerebralvalley.ai/u/azeem), [Ankita G](https://cerebralvalley.ai/u/gazubi)
- **GitHub:** https://github.com/nprabhakar/super-duper-fishstick
- **Demo video:** https://drive.google.com/drive/folders/1Un07vtOSGemOw9Aet-_zI_Sf-ZfR5Cqb?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=46

### 21. PropertySim

We built an AI tool that transforms ordinary property photos into an interactive 3D view and floor plan, helping users understand a space far better than flat images alone. By analyzing batches of room photos, Gemini develops a richer understanding of the layout, dimensions, openings, and furniture, while also powering guided tours through Gemini Live. The result is faster, smarter property discovery that reduces wasted visits and helps people find a place they love with more confidence.

- **Team:** [Reynaldo Wijaya](https://cerebralvalley.ai/u/rwhendry), [Steven Limanus](https://cerebralvalley.ai/u/slimanus), [Miki Miki](https://cerebralvalley.ai/u/ratrekt)
- **GitHub:** https://github.com/arielmiki/propertysim
- **Demo video:** https://youtu.be/S2hUupVABV8
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=52

### 22. https://youtu.be/2GR6-HRKaNI?si=WnqAq529bDHySRqv

BTO-Sensei is an AI-powered HDB (Housing & Development Board) inspection assistant. Built with React, TypeScript, and Vite, it uses the Gemini Live API alongside custom audio DSP to perform real-time acoustic analysis of floor tiles, capture visible construction defects via camera, and auto-generate structural integrity reports.
Eventual grand vision is for it to be integrated with google glasses with overlays!

- **Team:** [Wei Song](https://cerebralvalley.ai/u/leprekon)
- **GitHub:** https://github.com/leprekonsg/BTO-sensei
- **Demo video:** https://youtu.be/2GR6-HRKaNI?si=WnqAq529bDHySRqv
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=60

### 23. Draw and Fly

We don't always have our Nintendo Switch around, but we always have our phones with us. This gemini-powered game allows users to draw their map and render it in a unique map. They will be able to fly their plane using their phone as a controller to control the pitch and roll. The next step will be to introduce multiplayer features.

- **Team:** [Kuang Wen Chan](https://cerebralvalley.ai/u/kwen1510)
- **GitHub:** https://github.com/kwen1510/gemini-hackathon-070326
- **Demo video:** https://youtu.be/Y4Sh-cKP6yc
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=71

### 24. Berlayar

Escape the AI uses Gemini to understand a real location, generate a themed 3D escape room around it, create multimodal challenges, and react in real time as an AI Warden. Players complete voice, image, video, and social challenges in the real world, while Gemini verifies their actions and adapts the game live.

- **Team:** [Enjiao Chen](https://cerebralvalley.ai/u/erniesg)
- **GitHub:** https://github.com/erniesg/escapetheai
- **Demo video:** https://drive.google.com/file/d/1l0uSR92qXJOniyfgjUyLxwtWcsh4OhOT/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=29

### 25. rocket rides (Live satellite tracker)

Rocket Rides is a cinematic 3D space logistics game built with Google's Antigravity, where players pilot rockets between Earth cities, ISS delivery missions, and Moon operations (Yes, you can go to the moon. That takes more than 1 minute!)

The experience includes WebAR-powered hand gesture rocket control and navigation for immersive flight input beyond standard keyboard controls.

Our AI stack uses Gemini API for speech generation, speech-to-text mission interaction, and satellite analysis, while Nanobanana is used to generate custom Astronaut achievement badges.

Live satellite tracking is integrated through NASA API data so players can monitor and interact with active orbital traffic directly on the globe.

- **Team:** [Regan Tan](https://cerebralvalley.ai/u/regant), [reiwen ing](https://cerebralvalley.ai/u/r3iwen)
- **GitHub:** https://github.com/reiwening/rocket-ride
- **Demo video:** https://youtu.be/YS2gpmCsI0s
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=59

### 26. seakkx

Most housing decisions today are made from flat brochures and static floor plans — documents that can't communicate perspective, sunlight loss, or how neighbouring towers affect your daily view. Buyers are left guessing what a given floor actually sees, how surrounding developments shape openness and privacy, and where light disappears at different times of day. That spatial ambiguity is exactly where expensive housing decisions break down.
We close that gap by starting from the skyline, not the PDF. Upload a BTO brochure and the app composites product-specific geometry — blocks, floors, lift cores, unit orientations — onto a realistic city-scale 3D base layer. The result is a decision environment that is floor-aware, direction-aware, and context-aware, letting buyers compare street-level perspectives, simulate sunlight across seasons, and understand spatial trade-offs that no 2D document can convey.

- **Placement:** 3rd Place
- **Team:** [Kai Xun Juay](https://cerebralvalley.ai/u/kai_xun), [Sean Yap](https://cerebralvalley.ai/u/Lactrodectus), [kaelan wan](https://cerebralvalley.ai/u/kaenugget)
- **GitHub:** https://github.com/S3annnyyy/gemini-3-hackathon
- **Demo video:** https://www.youtube.com/watch?v=KaIytzGh_eU
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=37

### 27. HardlyWorking

HardlyWorking is a platform that lets anyone define, test, and deploy AI-powered agents that convert unstructured data (emails, documents, messages) into structured database records. Users visually design extraction schemas through a conversational interface powered by Google Gemini, backtest them against real samples, and deploy production agents that listen on a dedicated email endpoint and persist extracted data to Supabase.
The core insight: most teams sit on valuable unstructured data (RFQs, invoices, support tickets, partner communications) but lack the engineering resources to build extraction pipelines. HardlyWorking collapses that process from weeks of development into a single interactive session.

- **Team:** [Lye junshen](https://cerebralvalley.ai/u/captainmayhem), [Franz Friedel](https://cerebralvalley.ai/u/FranzFriedel), [Ching Yen](https://cerebralvalley.ai/u/Ching)
- **GitHub:** https://github.com/junshenlye/Gemini3_Deepmind.git
- **Demo video:** https://youtu.be/UVmxgxZfwKQ
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=58

### 28. WildCard

WildCard turns Singapore's extraordinary biodiversity into a collectible card game you play in the real world. Singapore is home to over 400 bird species and remarkable urban wildlife, yet most residents walk right past it. Existing biodiversity tools are scientifically valuable but offer no engaging consumer loop.
WildCard fixes this. Photograph any animal in the wild, and the app identifies the species, surfaces educational facts, and generates a unique 1-of-1 stylised collector card. Each card carries stat ratings, a rarity tier derived from real GBIF Singapore sighting data and the Singapore Red Data Book, and a sighting stamp with location and date. Rare, Epic, and Legendary cards unlock a short animation, bringing the illustration to life. Cards are saved to a personal field guide tied to Singapore's complete species registry (~300 fauna records).

- **Team:** [Justin Wong](https://cerebralvalley.ai/u/juswwj)
- **GitHub:** https://github.com/juswwj/gemini_hack
- **Demo video:** https://youtube.com/shorts/YEWK7r_y7OU?feature=share
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=67

### 29. VASS

Project VASS (Voice-to-Action Logistics & Sentiment Engine)

- **Team:** [Sakthiish Prince](https://cerebralvalley.ai/u/Sakthiish), [Musang Musang](https://cerebralvalley.ai/u/Musang), [Steven Teng](https://cerebralvalley.ai/u/ferryman), [Valencia Q](https://cerebralvalley.ai/u/val-valsea)
- **GitHub:** https://github.com/sakthiish12/HackathonMarch9
- **Demo video:** https://valsea-clone-buddy.vercel.app/
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=70

### 30. Krashless

Problem: Singapore has bad pedestrians and ever worse drivers.

Solution: Krashless is an agentic urban prototyping engine for transport authorities like LTA, shifting road safety from reactive to proactive.

Using 3D maps and live CCTV, Krashless visualizes high-risk hotspots and automatically generates infrastructure blueprints. It visually overlays physical changes, like bollards, alongside costs and risk justifications.

Crucially, it features a human-in-the-loop "Review Panel". If an engineer rejects an AI proposal (e.g., "bollards block double-decker turning radii"), our Explainable AI instantly generates an adapted alternative on the fly. Krashless helps design safer cities, faster.

- **Placement:** 2nd Place
- **Team:** [Fong Kai JIe](https://cerebralvalley.ai/u/KaiJie), [adore blvnk](https://cerebralvalley.ai/u/adore_blvnk)
- **GitHub:** https://github.com/adoreblvnk/krashless
- **Demo video:** https://www.youtube.com/watch?v=UD9xvQjf4XQ
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=72

### 31. Cannot Decide

The "Decision Buddy" project aims at helping indecisive people go through every single decision from choosing their dinner, whether to date someone, and even choosing from their job offers. Powered by Gemini 3 Pro with High thinking level, it can be personalized based on each user's preferences and life experiences.

- **Team:** [HSIN-YUN HAN](https://cerebralvalley.ai/u/Hiiian)
- **GitHub:** https://github.com/Quackqqq/decision-buddy
- **Demo video:** https://youtu.be/yGmy4G4Jsc4
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=38

### 32. 🌱 Canopy Art & Wine

🌱Canopy Art & Wine is a "Google Maps with Smart Routing for Art & Wine," a Routing Oracle that combines 18M+ offline auction records, real-time blockchain data (22 chains), and $27T+ in stablecoin intelligence. It helps Art & Wine enthusiasts their assets with speed, confidence, and the right price.

This prescriptive layer for the $90B market provides agents with turn-by-turn directions for transactions, reducing a 3-day, $4,200 wire transfer to $2.30 settled in 47 seconds. 

The complete protocol stack (ERC-8004, A2A, x402, AP2) forms a cryptographically verified execution loop, transforming Canopy from a data product into autonomous infrastructure.

- **Team:** [Khanh Dang](https://cerebralvalley.ai/u/khanhfromla)
- **GitHub:** https://github.com/calxcape/CanopyArtWineAGHackathon
- **Demo video:** https://drive.google.com/file/d/18inadZppLED598Tgoh47Z2Ld4rNGUwS2/view?usp=drive_link
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=40

### 33. shun

Workflow Buddy is a Chrome extension that turns a live browser walkthrough into an AI-ready workflow spec. Instead of asking users to manually document every step, it captures clicks, typed inputs, page context, and screenshots in a structured format that another LLM can use to generate reliable browser automation. It solves the gap between “I can show the task” and “an AI can reproduce it,” making workflow automation faster, clearer, and far less error-prone.

- **Team:** [shun shun](https://cerebralvalley.ai/u/shunshun)
- **GitHub:** https://github.com/tanshunyuan/workflow-buddy
- **Demo video:** https://youtu.be/JanW73LZ5-Y
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=41

### 34. DefaultTaste

Agent Taste Profiling as a Service!!
- **DefaultTaste** is a platform that reveals the hidden aesthetic preferences of AI agents.
- It works by probing an agent repeatedly across many generations, then measuring the patterns it keeps returning to, such as colors, fonts, layouts, genres, tempo, and mood.
- The result is a clear **taste profile** plus a **correction prompt** that helps users override those defaults.
- The problem it solves is that most people judge AI from a single output, which hides the model’s underlying bias and stylistic tendencies.
- In practice, this means creators often ship outputs shaped more by the model’s default taste than their own intent.
- DefaultTaste makes those defaults visible and actionable, so developers, designers, and creators can steer AI outputs toward something more original, intentional, and differentiated.

- **Placement:** 1st Place
- **Team:** [Ilham Firdausi Putra](https://cerebralvalley.ai/u/ilhamfp), [Aufa Husen](https://cerebralvalley.ai/u/aufahr)
- **GitHub:** https://github.com/ilhamfp/defaulttaste
- **Demo video:** https://www.loom.com/share/c13cd31350cb446994fc167ecc96918b
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=44

### 35. GemScrop AI-Powered Market Intelligence for stock

GemScope is a real-time market intelligence platform powered by Google Gemini that runs 24/7 autonomously. Multiple specialized AI agents work through a 6-phase pipeline — gathering intel from news, YouTube, and social media, fact-checking findings, performing deep analysis, and generating actionable stock recommendations with quality scoring. The AI autonomously selects trending topics and researches them non-stop without human intervention. Users can watch agents think and collaborate in real-time, or submit their own topics for priority processing. Built with Next.js, .NET, PostgreSQL, and Vertex AI.

- **Team:** [Zaifu Zheng](https://cerebralvalley.ai/u/richagain)
- **GitHub:** https://github.com/richagain/h26zaifu
- **Demo video:** https://www.youtube.com/watch?v=4upa1lBpPi4
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=47

### 36. PedalSG

The project imports data for all cycling paths and parking spots from LTA, let's the user figure out a cycling path from point A to point B depending on what kind of path they want (safe,scenic, sheltered etc), and then simulate the kind of view they are expected to see based on the coordinates of the path, timing and weather. I use nanobanana to enhance to generate the images by sending in the geo-coordinates + the weather fetch + the time.

- **Team:** [Satvik Sinha](https://cerebralvalley.ai/u/satviksinha)
- **GitHub:** https://github.com/satviksinha/PedalSG
- **Demo video:** https://youtu.be/m5ahOsPMkho
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=54

### 37. OffBeat

Offbeat is a smart, context-aware travel companion that redefines how you spend the "in-between" moments of your day. Rather than focusing on big itineraries, Offbeat focuses on the gaps—the hour between meetings, the sudden rain delay, or the spontaneous urge for a quick walk.

- **Team:** [Ishan Mittal](https://cerebralvalley.ai/u/ishan20mittal), [Vamsi Grandhi](https://cerebralvalley.ai/u/vamsig), [Mahdi Abolfazli Esfahani](https://cerebralvalley.ai/u/mahdiabol), [Venkataraman Natarajan Iyer](https://cerebralvalley.ai/u/survivelifeform)
- **GitHub:** https://github.com/maesfahani/Offbeat
- **Demo video:** https://drive.google.com/file/d/1f5U3FMRRQrbt8lYyqxqKyAZzmlWJxffc/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=61

### 38. SingFlix

Turn any Singapore landmark into an instant immersive experience!

Singlix lets user upload a photo of a real Singapore attraction and instantly detects the location name, coordinates, and generates a short cinematic video that brings the place’s history and cultural significance to life based on actual historical information and urban planning data. Instead of reading plaques or searching online, users can experience the story of the place in a visually rich, engaging format right where they are.

- **Team:** [Esther Wang](https://cerebralvalley.ai/u/eswyl17), [Weng Siong Chan](https://cerebralvalley.ai/u/weng_siong)
- **GitHub:** https://github.com/Dylon-Chan/gemini3-hackathon-sg
- **Demo video:** https://www.loom.com/share/7341f895410742319dab7fdcb3bdd50b
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=63

### 39. polyglot

ASL-lingo broadens and accelerates American Sign Language adoption by closing the learning loop. Traditional tools are one-directional: they show signs but never verify if the learner is signing correctly. ASL-lingo lets users learn a sign, perform it on camera, and receive real-time feedback.
                                                                                                                                                                                                                                                                                                                                                Tech stack: Next.js · React · TypeScript · Tailwind CSS · MediaPipe Tasks Vision · Google Gemini API · Git

- **Team:** [Ravi Vats](https://cerebralvalley.ai/u/ravivats), [Anh Ha](https://cerebralvalley.ai/u/hahaqanh), [Siva Valliappan](https://cerebralvalley.ai/u/SivaInSG), [Andrew Soon](https://cerebralvalley.ai/u/andrewsoon)
- **GitHub:** https://github.com/siva73-git/learning-asl
- **Demo video:** https://drive.google.com/drive/folders/1IY-pXmIRyzeYkWCKc7uomwTLhC2jwMrw?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=65

### 40. Edmund Space Solutions

ESS Airlock Auditor: Autonomous Engineering Oversight

We have built a Reasoning Agent that transforms static NASA Technical Standards into a live safety auditor. Using Gemini 3 Flash's native multimodality, our system analyzes hardware telemetry against 1,000+ pages of NASA-STD-3001, 5001, and 7012 simultaneously.

- **Team:** [Edmund Anthony](https://cerebralvalley.ai/u/Edmundzen)
- **GitHub:** https://github.com/edmundzen/Airlock-Auditor-Gemini3
- **Demo video:** https://youtu.be/b-akxglDJww?si=CK3Fz7HthiGZoIB6
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=66

### 41. Fooyo

Creating high-quality video content typically requires professional skills, expensive production teams, and complex tools for animation, camera control, and editing. This makes storytelling through video difficult for marketers, educators, and cultural institutions who want to communicate ideas quickly and visually.

AI Director solves this problem by turning video production into an intuitive directing process. Using natural language and visual references, users can place characters and objects within scenes, control 3D human motion and poses, and position virtual cameras just like a film director. The system translates ideas into staged visual scenes, enabling anyone to produce cinematic storytelling content without technical expertise or large production budgets.

- **Team:** [Shaohuan Li](https://cerebralvalley.ai/u/Shaohuan), [Tianjun Zang](https://cerebralvalley.ai/u/Bruce_ZANG), [James Wang](https://cerebralvalley.ai/u/wwjjames)
- **GitHub:** https://github.com/Lincolnnus/gemini-hackathon
- **Demo video:** https://youtu.be/3tvrfT6criQ
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=68

### 42. openME

A Go-based RAG orchestrator for local LLMs, integrating SearXNG and Qdrant via MCP.

- **Team:** [T Pughal Raj](https://cerebralvalley.ai/u/tamago55)
- **GitHub:** https://github.com/Pughal77/llm-compiler
- **Demo video:** https://youtu.be/JKfXPK-XF1k?si=9Of5VhR7ItCUXOBO
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=73

### 43. Echoverse

Every sound has a universe hiding in it.

Echoverse records ambient sounds and generates fictional universes where those sounds are native. A coffee shop's clinking cups become crystalline chime-beetles in an underground bazaar. Traffic hum becomes the breath of colossal stone guardians.

Built entirely on Google AI — Gemini for imagination, Imagen 4 for visuals, Veo 3.1 for video.

- **Team:** [Vishal Banthia](https://cerebralvalley.ai/u/vbanthia)
- **GitHub:** https://github.com/vbanthia/echoverse
- **Demo video:** https://www.youtube.com/watch?v=_-Q0zVN3XQM
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=91

### 44. Way too Claude

**Evil Twin interrogates you for 2 minutes — your projects, your GitHub, your ambitions, your excuses — and builds a complete personality model. Then it generates your City of Procrastination: a 3D city where every building is something you've been putting off, sized by how long you've been avoiding it. Your Evil Twin — a moustached, zero-hesitation clone of you — enters the city and starts conquering every building you haven't, shipping your landing page, emailing your investors, messaging your cofounder, all in your voice, your style, your technical taste, but without the fear. You stand at the entrance and watch. Then you see everyone else's cities — your friends, your team, your cohort — and when your procrastination is public and an AI is out-executing you in front of everyone, you stop making excuses. High agency, enforced by community.**

- **Team:** [Arnav Jhajharia](https://cerebralvalley.ai/u/arnavjhajharia), [Vangmay Sachan](https://cerebralvalley.ai/u/Vangmayy)
- **GitHub:** https://github.com/Arnav-Jhajharia/evil-twin
- **Demo video:** https://youtu.be/RqIz3ffKeSs
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=95

### 45. Nexus

Nexus is a personal AI companion and social intelligence platform that helps users find the right people, opportunities, and communities for their next chapter in life. Through ongoing conversation, uploaded documents, memory, and user-controlled preferences, Nexus builds a deeper understanding of a user’s goals, transitions, skills, concerns, and personality. It then uses a multi-agent system to explore a living network of agents and surface meaningful matches such as mentors, collaborators, communities, and career connections.

The platform includes three core experiences. First, a premium chat interface where users speak naturally with their AI companion. Second, a dashboard where users can inspect and edit memory, upload documents, manage extracted knowledge, and personalise the AI’s behaviour. Third, a dynamic world visualisation where users can see their agent exploring clusters, communicating with other agents, and forming introductions in real time.

- **Team:** [Bernardino Lintang](https://cerebralvalley.ai/u/bernard_ino), [Jia Chin Ng](https://cerebralvalley.ai/u/jainge), [Carlos Chacon](https://cerebralvalley.ai/u/cchakons), [Dhasarathan Deepan Krishnaa](https://cerebralvalley.ai/u/deepankrishnaa)
- **GitHub:** https://github.com/cchacons/gemini3_hackathon_sg
- **Demo video:** https://youtu.be/CDJ_mJ6KgUo?si=LyBWr3edbbDV2mFp
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=98

### 46. izan

izan leverages Gemini 3 to create an autonomous, self-healing "agen" that replaces high-priced SaaS with low-cost open-source infrastructure, effectively reducing cost by saas by 10 times.

- **Team:** [Kunal Gulati](https://cerebralvalley.ai/u/Alllalalalalala)
- **GitHub:** https://github.com/agentic-k/izanagi, https://github.com/agentic-k/infrastructure, https://github.com/agentic-k/gemini-izan-portal
- **Demo video:** https://app.canvid.com/share/fi_01KK3S7M7YJWZS43RZ8ZSAH76J
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=99

### 47. PLQHeroes

An open-source interactive television application designed to transform passive screen time into an active educational experience. By seamlessly blending parent-curated video content with real-time activities and quizzes, KangarooTV ensures that children aren't just watching—they’re participating.

- **Placement:** Finalist
- **Team:** [Juan Miguel Jimeno](https://cerebralvalley.ai/u/jimenojuan), [Hasala Marakkalage](https://cerebralvalley.ai/u/hasalams), [Josey Mathew](https://cerebralvalley.ai/u/tijm), [Kuruvachan Kalluvelil George](https://cerebralvalley.ai/u/KKGeorge)
- **GitHub:** https://github.com/tijoseymathew/KangarooTV
- **Demo video:** https://www.youtube.com/watch?v=S0x3Of47T28
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=101

### 48. Checkout court

Checkout court is a browser extension that puts your online purchases on trial. When you click checkout, the extension blocks the button and summons you to a short AI-powered courtroom where you must defend the purchase.

The AI knows your savings targets, spending limits, and progress toward your goals, and uses that information as evidence during the trial. Acting as judge and prosecutor, it questions whether the purchase fits your financial plan, challenges weak justifications, and weighs the cost against what you’re trying to save for.

If your case holds up, the checkout is unlocked. If not, the verdict is to reconsider. By turning checkout into a quick courtroom debate, Checkout Court helps ensure every purchase stays aligned with your long-term financial goals.

- **Team:** [Alif Daffa](https://cerebralvalley.ai/u/Coder084), [Aqil Akram Ghazali](https://cerebralvalley.ai/u/rubyonline), [Shafiq Ninaba](https://cerebralvalley.ai/u/shafiqninaba)
- **GitHub:** https://github.com/shafiqninaba/checkout-court
- **Demo video:** https://drive.google.com/drive/folders/1ssycSRVi0_hBO20RXk54vhzkY389n9AA?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=74

### 49. geminiceguys

## IHEE — Iterative Hypothesis Elimination Engine

Financial crises don't announce themselves. IHEE is an autonomous risk investigation system powered by Gemini 2.5 Flash that turns market signals into structured diagnoses — in minutes, not days.

When a signal is detected (CDS spike, equity drop, credit event), IHEE generates competing hypotheses and systematically eliminates them using evidence — exactly like a forensic analyst, but at machine speed.

- **Team:** [Saai Aravindh Raja](https://cerebralvalley.ai/u/SaaiAravindhRaja), [Aarnesh Khaitan](https://cerebralvalley.ai/u/aarnesh), [Vatsal Parikh](https://cerebralvalley.ai/u/spiffler)
- **GitHub:** https://github.com/AarneshKhaitan/GeminiHackathon
- **Demo video:** https://www.youtube.com/
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=80

### 50. steffenstein

# Repo Rescue — Hackathon Description

     ## What it does

     Repo Rescue uses Gemini to find the unsung heroes in any GitHub repository. You
     enter a repo, and an AI agent automatically scans for critical bugs, security
     vulnerabilities, and incidents — then traces who fixed them, how fast they
     responded, and how often they worked late nights to keep the project alive. The
     result is a ranked leaderboard of "firefighters" with scored hero cards.

     ## The problem

     Open source maintainers burn out. Contributors who quietly fix the hardest bugs —
     patching CVEs at midnight, triaging crashes on weekends — rarely get recognized.
     There's no easy way to see who actually keeps a project running. GitHub shows
     commits and PRs, but not who consistently shows up when things are on fire.

- **Team:** [Steffen Stein](https://cerebralvalley.ai/u/steffen)
- **GitHub:** https://github.com/therealstein/gemini-reporescue-quest
- **Demo video:** https://www.loom.com/share/9efe488ca2ff4681bc355f64fd089aa3
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=81

### 51. NeuralForge

NeuralForge is a Gemini-powered workflow that turns machine learning models into silicon.

You describe a neural network in plain English, and NeuralForge converts it into synthesizable Verilog, embedding the trained weights directly into hardware logic.

The system then runs a real RTL-to-GDS pipeline,  synthesising with Yosys, verifying functionality through simulation, and implementing the design on the SkyWater 130nm PDK using open-source backend tools.

AI agents monitor tool outputs, fix design errors, and iterate until the hardware is physically valid.

The model isn’t deployed on hardware.
It becomes the hardware.

- **Team:** [Jishnu Venugopal](https://cerebralvalley.ai/u/jishv), [Jibin Mathew](https://cerebralvalley.ai/u/jibinmathew)
- **GitHub:** https://github.com/jishnuvenugopal/neuralforge
- **Demo video:** https://drive.google.com/drive/folders/1vgXg2_nQBC1YnCR2MaWuzsYtZE-DqHzk?usp=drive_link
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=82

### 52. PathMakers

NaviSwarm is an AI-powered spatial intelligence assistant that runs directly in your mobile browser. It uses the Gemini Multimodal API to autonomously map your physical environment in real-time, providing vision-impaired or lost users with augmented reality (AR) guidance, natural conversational directions, and live hazard detection—all without requiring GPS or dedicated hardware.

- **Team:** [Lin Zhi](https://cerebralvalley.ai/u/linzhi), [jimmy Chiun](https://cerebralvalley.ai/u/jimmy291), [Navin Kaviyarasu](https://cerebralvalley.ai/u/navin_kaviyarasu)
- **GitHub:** https://github.com/Jimmy291/Gemini3Hackathon
- **Demo video:** https://drive.google.com/file/d/12zxyG7QdYmeKKukI666gXoGxLIZl1_Yq/view?usp=share_link
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=75

### 53. Daisy Gravity

A page where you can seek the answer for all kind of problem, just not very credible...

- **Team:** [Nick Pat](https://cerebralvalley.ai/u/nickpat)
- **GitHub:** https://github.com/nickpat37/Daisy-gravity.git
- **Demo video:** https://youtu.be/HnEHC94koBA
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=77

### 54. UTAMA.AI

Utama AI — Step into Singapore's past through the lens of your iPhone. Point your camera at any flat surface and watch Sultan Hussein and his guardian lion materialize before your eyes in stunning AR. Have a real-time voice conversation with the Sultan — powered by Google's Gemini Live AI — as he recounts the founding of modern Singapore, while his lion prowls, roars, and even triggers an immersive VR flashback sequence. Built for the National Heritage Board, Utama AI transforms static history into a living, breathing encounter you can talk to.

- **Team:** [Jing Wen Zhang](https://cerebralvalley.ai/u/juno)
- **GitHub:** https://github.com/pebblepaw/UTAMA_AI
- **Demo video:** https://youtube.com/shorts/f3ol2L_EAtY?feature=share
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=100

### 55. Wei Jun #1

During bbarbeque, dinners, and night out with friends, a little karoake is something anyone can bond over. But what if some of your frieds don't know how to sing!?  We can all have a little fun together, and use AI to make any night enjoyable

This Karoake Roaster is a chaotic multiplayer web application, where everyone is singing and being graded by a brutal judge persona, Gemini 2.5 Flash, who reads our singing abilities in the form of personas.  It checks for pitch, tonality, rhythm, and stability, and it is critical with grading you!

A fun part is that Gemini uses your friends agianst you. Before the singing starts, anyone else will write an assumption of how well you perform. Gemini will use those assumptions and roast the singer with it. Everyone can see who wrote what!

And the personally to me, more funny part, is that your webcam will be activated, and everyone can see you sing in real time, making it so even you and your friends can play together far away from each other.

- **Team:** [Ang Wei Jun](https://cerebralvalley.ai/u/gnajun17)
- **GitHub:** https://github.com/angweij1431/Cerebal-hackathon-karoake-roaster
- **Demo video:** https://drive.google.com/drive/folders/1CLPLvkBfqupxY2apwtLMla7R3vJlFpfH?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=76

### 56. Chris AI producer

This is a Ai film production producer app

- **Team:** [Kun Yan](https://cerebralvalley.ai/u/ChrisJoJo)
- **GitHub:** https://github.com/kunkunkunkun/ChrisAiFilmProducer
- **Demo video:** https://youtu.be/U13T4C4cYWU
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=78

### 57. Prism

Togetherly is a communal AI-powered card platform that transforms a group's individual messages into a single cinematic celebration — an evolving visual artwork paired with an original AI-generated song, delivered as a full-screen reveal that feels like a music video made entirely about you.

The problem it solves: creating a personalised group card is disproportionately hard, and even when it exists, the experience ends the moment it's received. One person carries all the coordination burden. The recipient reads it once and it's forgotten.

Togetherly eliminates both problems. The creator sets up in under 2 minutes. Contributors join via a single link — no account, no app — and each message evolves a shared AI-generated visual in real time. When the deadline closes, Gemini synthesises every message into a unified statement and generates an original song from it.

- **Team:** [Jieying Zeng](https://cerebralvalley.ai/u/jieying)
- **GitHub:** https://github.com/IncommensurableHubris/gemini-hackathon-living-card.git
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=86

### 58. AIOS

AIOS is the first OS where AI isn't an app — it's infrastructure.

Problem: On every OS today, AI lives in a chat window. The OS itself has no idea what you're doing, can't search files by meaning, and has zero security model for AI agents.

Solution: A Linux-based OS with:

• Intent routing — say what you want, the OS figures out how
• Semantic filesystem — search files by meaning, not filename
• Always-on local LLM — every process can call AI like a system API
• Zero-trust agent security — kernel-enforced crypto capability tokens
• Self-healing — detects, diagnoses, and fixes issues autonomously
• Tamper-evident audit — hash-chained logs for every action

Fully local. Fully open-source. Zero cloud dependency.

- **Team:** [sanford ng](https://cerebralvalley.ai/u/mai_loh), [Nathaniel Neo](https://cerebralvalley.ai/u/Nattyyy), [Jia Han Kia](https://cerebralvalley.ai/u/jiahan), [ainsley woo](https://cerebralvalley.ai/u/ains_tein)
- **GitHub:** https://github.com/Kantosaurus/some-google-shii
- **Demo video:** https://youtu.be/KHVi2Kzn5gk
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=88

### 59. Patched

A chat widget which let's developers make their apps user-editable in real time.

Step 1: Developer of a SaaS/WebApp adds our widget package to their nextJS app.
Step 2: Developer deployes a backend service that uses Gemini CLI with flash models.
Step 3: Users type in a feature request/UI change/customization using the widget, and it gets deployed in realtime (~1 min).

- **Team:** [Rohan Sood](https://cerebralvalley.ai/u/rohansood15), [lionel g](https://cerebralvalley.ai/u/loinel), [Arpit R](https://cerebralvalley.ai/u/whoisarpit)
- **GitHub:** https://github.com/patched-codes/gemini-hackathon-submission
- **Demo video:** https://youtu.be/HjloUtuMggM
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=90

### 60. TUANG - Take a Day Off

TUANG your life

- **Team:** [Gwendalynn Lim](https://cerebralvalley.ai/u/1ightray)
- **GitHub:** https://github.com/evecount/tuang_takethedayoff
- **Demo video:** https://youtu.be/3I6FtBBAeio
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=79

### 61. Red Cat Labs

YouTube Channel Strategist : Analyses a YouTube channel's videos (with leave-one-out approach), to deliver estimates for existing and new videos, along with a rubric about dos-and-dont's

- **Team:** [Martin Andrews](https://cerebralvalley.ai/u/mdda)
- **GitHub:** https://github.com/mdda/YouTubeStrategist
- **Demo video:** https://drive.google.com/file/d/1w0usZLHfczHexEn-3xhvMpRaxdfT-f3Z/view
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=84

### 62. Lemondae

“Every great book has a moment that stops you cold. A scene so vivid, so emotional, you can feel it. What if your screen could feel it too?” Book2Life brings your book to life by generating a mini film based on any passage or text , generates soundtracks with different moods - character, cinematic and reader. It also gives an analysis of those scene for those interested in more detailed nuances!

- **Team:** [rhea s](https://cerebralvalley.ai/u/rheaa11)
- **GitHub:** https://github.com/rheaaas11/book2life
- **Demo video:** https://drive.google.com/file/d/1MbKCNOUbIkqYgcEf1MIcnBu6o3bfrysP/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=83

### 63. impromptu

A party game where AI (and human) collaborate to tell a story one word at a time, improv style.

- **Team:** [cc ho](https://cerebralvalley.ai/u/ccho)
- **GitHub:** https://github.com/stranger26/impromptu
- **Demo video:** https://youtu.be/G9mHZFSh5Ek
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=85

### 64. sgpulsifier

SG Pulse

- **Team:** [Priyesh Vishwakarma](https://cerebralvalley.ai/u/vkpriyesh)
- **GitHub:** https://github.com/vkpriyesh/sgpulse-gh-26
- **Demo video:** https://youtu.be/8hwJn7YWq0o
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=87

### 65. Continual Learning Interface (CLI)

We propose a thought experiment, what if a machine had a continual learning mechanism? 
If machines are exceeding human intelligence, machines should be improving themselves, not humans
This way, humans, no longer bottleneck the ingenuity of machine intelligence

This project proposes a proof of concept of what could be a continual learning interface CLI. 
We demonstrate that a Browser Agent could be given very specific human feedback and improve itself, whether that's through updating or creating new scripts in its own system. 
And it does so by a meta-agent or a meta-system that updates the orchestrator agent (which is itself).

- **Team:** [Brandon Ong](https://cerebralvalley.ai/u/ob1)
- **GitHub:** https://github.com/theogbrand/godel-darwin-machines
- **Demo video:** https://youtu.be/RNpbSHINLug
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=89

### 66. Charles Yap

TaskStream is a voice-driven project management app for iOS that acts as an intelligent personal assistant for busy professionals juggling multiple projects. 

Instead of manually sorting through emails, calendars, and task lists across different tools, users simply open the app and ask for a "board update" — the AI scans their Gmail inbox, Google Calendar, and Drive, then synthesizes everything into a prioritized summary with actionable suggestions like creating projects, assigning tasks, or scheduling meetings. 

It solves the daily overhead of context-switching and information triage: rather than spending 30 minutes each morning reading emails and updating spreadsheets, the app reads everything for you, figures out what matters, and lets you approve actions with a single tap.

- **Team:** [Charles Yap](https://cerebralvalley.ai/u/charlesyapai)
- **GitHub:** https://github.com/charlesyapai/PM-bot-deepmind-hackathon
- **Demo video:** https://youtu.be/uncuwle5s24
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=92

### 67. Glint

The Essence:

"An AI-orchestrated intelligence engine that deconstructs global multi-language narratives into structured, cross-verified insights for everyone."

The Problem Solved:

"We bridge the gap between expensive, niche-specific intelligence software and the average user’s inability to unlock the full analytical potential of Large Language Models."

- **Team:** [Wei Qin](https://cerebralvalley.ai/u/Veetsin), [manman zeng](https://cerebralvalley.ai/u/miniSuperman)
- **GitHub:** https://github.com/veetsin/Singapore-Hackathon
- **Demo video:** https://drive.google.com/file/d/1Ua5dkF21stDlULCTaJ8omvWW26OcSND4/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=93

### 68. cues

AutoSOP leverages Google Gemini's multimodal vision to instantly watch your screen recording and automatically generate a beautiful, editable, step-by-step Standard Operating Procedure in seconds.

- **Team:** [Deric Zeng](https://cerebralvalley.ai/u/deric)
- **GitHub:** https://github.com/waynnbz/g3-hack-autosop
- **Demo video:** https://youtu.be/vuc1XfJUMDs
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=94

### 69. MySpec

An AI agent that bridges the gap between vibe coding and spec-driven development. It turns your idea into executable documentation. It asks the right questions to produce a “Spec Bundle” — clear, detailed Requirements, Design, and Task files — that can be fed directly into Gemini CLI, Google AI Studio, or other vibe-coding tools to build with far more precision so your coding agent works from a clean source of truth instead of guesswork.

- **Team:** [Vincent Tran](https://cerebralvalley.ai/u/0xgosu)
- **GitHub:** https://github.com/0xGosu/gemini-hackathon-project
- **Demo video:** https://youtu.be/h9TOeJeEbmo
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=96

### 70. Serena

Most people have hundreds of videos sitting on their phone that never get posted. Not because the moments aren't good -  but because editing takes forever and most people don't know where to start.
Supa fixes that. It goes through your footage, finds the best clips, edits them together, and adds your voice. You go from a camera roll full of videos to a TikTok ready to post, without touching a timeline.

- **Team:** [Serena Lam](https://cerebralvalley.ai/u/Serenalam)
- **GitHub:** https://github.com/sslam3
- **Demo video:** https://screen.studio/share/Oj0rGjZl?state=uploading
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=97

### 71. JH

GenomePuzzle-Interactive Genome Education Platform

- **Team:** [JiaHao Tan](https://cerebralvalley.ai/u/jiahaotan)
- **GitHub:** https://github.com/jiahao1553/hackathon
- **Demo video:** https://youtu.be/2OTXPRMGALU
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=102

### 72. i made it

Skip manually clicking of buttons on ecommerce sites with this agent

- **Team:** [Kai Ming Gan](https://cerebralvalley.ai/u/kmgannn)
- **GitHub:** https://github.com/kmgannn/gemini-hackathon-070326
- **Demo video:** https://youtu.be/nSCtXkU5kdo
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=103

### 73. AdmitPlus.AI

AdmitPlus is building the AI Operating System for global students.

Most AI education tools are stateless assistants: students get feedback, close the tab, and the system forgets everything.

But learning is longitudinal — improvement requires memory, adaptation, and continuous feedback.

AdmitPlus introduces a persistent AI learning system for students preparing for exams like IELTS, TOEFL, and university admissions.

Powered by Gemini-based multi-agent architecture, the platform coordinates specialized agents in a continuous learning loop:
Task Agent generates targeted exercises,
Evaluation Agent scores responses using rubric-based reasoning,
Feedback Agent delivers structured improvement guidance, and
Planner Agent updates personalized study plans.

All interactions are written into a longitudinal memory layer, creating a dynamic academic profile for each student.

AdmitPlus transforms AI tutoring from isolated responses into a persistent intelligence layer for global education.

- **Team:** [Hao Jia](https://cerebralvalley.ai/u/meethao)
- **GitHub:** https://github.com/admitplus/admitplus-google-gemini
- **Demo video:** https://youtu.be/Hk7dcYE6GU4
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=1

### 74. Musical Movers

https://musical-mover.web.app/  (check it out!)

Musical movers help uncoordinated people move to the beat of the music. 

Users just have to tap to the beat or use their microphone to input the beat of the music, and they can copy the movement in their walks, runs or on the dance floor.

- **Team:** [Kevin Wang](https://cerebralvalley.ai/u/kevinishness)
- **GitHub:** https://github.com/caveman-sg/musicmover
- **Demo video:** https://www.loom.com/share/39766709ff3e49a5bd0a993efcb5e994
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=5

### 75. SplitReceipt

SplitReceipt is a local-first shared expense coordination app built for real group spending, not just hackathon screenshots. It turns a messy receipt into a shared expense, computes who owes whom, generates the minimum set of settlement transfers, and lets users confirm repayments through direct actions or natural-language prompts.

SplitReceipt targets a recurring, real-world pain point: shared spending between roommates, travel groups, event organizers, and small teams. The core value goes beyond the hackathon because it reduces both the arithmetic burden and the coordination burden of paying people back.

- **Team:** [Kuan-Hao Chen](https://cerebralvalley.ai/u/nghdavid), [Akash Hegde](https://cerebralvalley.ai/u/akashhegde2012)
- **GitHub:** https://github.com/akashhegde2012/split-ai
- **Demo video:** https://www.youtube.com/watch?v=hTMiqGiAeXQ
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=6

### 76. poor man hedge fund

Trading the Narrative: A POC of an AI Native Hedge Fund

An AI-native macro hedge fund prototype, inspired by YC's request for startups (https://www.ycombinator.com/rfs - AI native hedge fund). This system uses Large Language Models as a core reasoning and synthesis layer to bridge the gap between qualitative narratives and quantitative execution.

To be AI-native, I focused on leveraging what LLMs are actually good at: (1) historical extrapolation (2) qualitative extraction and analysis (3) reasoning around text. 

We leave the quantitative heavy lifting to existing methodologies via code, and focus on making a good financial qualitative reasoning engine.

- **Team:** [Yi Peng Neo](https://cerebralvalley.ai/u/Nyp)
- **GitHub:** https://github.com/neoyipeng2018/poor-man-ai-hedge-fund
- **Demo video:** https://youtu.be/LugRJ4Ak4kw
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=2

### 77. HTMM

JoNino is a multi-agent AI trading command centre that solves information overload for day traders. It consolidates live earnings calls, stock charts, SEC filings, and news into a Bloomberg-style tiled workspace where six Gemini-powered AI agents watch, analyse, and act in parallel.

Set rules in plain English — "if sentiment turns negative about Intel, set a stop-loss" or "if double top on AAPL, sell" — and the system monitors all sources autonomously. Gemini Live transcribes video in real-time, Gemini Pro analyses SEC filings and chart screenshots via vision, and Gemini Flash scores news sentiment. When a rule fires, an action card appears with reasoning and confidence. Execute trades to Alpaca paper trading manually or on auto.

Every trade flows through a learning loop: a reflection agent classifies decisions using Van Tharp's process-outcome matrix and extracts lessons that tune future thresholds — so the system improves with every trade it makes.

- **Team:** [Brian Isaac](https://cerebralvalley.ai/u/BrianIsaac)
- **GitHub:** https://github.com/BrianIsaac/jonino
- **Demo video:** https://youtu.be/SYbmdHXjNwQ
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=3

### 78. Bandung Panas with Extra Ice

Vibe Robotics turns plain English into a complete cardboard robot build using only components you already own.

Non-technical teams can't translate ideas into hardware, losing credibility with vendors and clients. ChatGPT and Claude make it worse. LLMs assume ideal parts and send you shopping, killing momentum before a single wire is connected.

Upload your inventory, describe what you want, and get a step-by-step slideshow, wiring diagram, and firmware code in under 5 minutes. Hit a problem? Upload a photo and it debugs your specific build on the spot.

This is not a chatbot wrapper. It generates a coherent build package constrained by your actual inventory.

Validated today (7 Mar 2026, 3:41pm) by an EdTech hardware company whose non-technical staff need physical prototypes to pitch vendors and clients.

Next: Digi-Key integration for auto-procurement of missing parts.

- **Team:** [Chang Yao](https://cerebralvalley.ai/u/justcy)
- **GitHub:** https://github.com/ChaiChangYao/hack
- **Demo video:** https://youtu.be/0JfqpzZB_sc
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=4

### 79. Gemini Craft

Gemini Craft is a Roblox/MineCraft style app that allows users to create a voxel style avatar to play in games. For this hackathon, there are 2 prebuilt mini games. Users can also use Gemini models to vibe create new experiences within the app. Different models can be used for different results.
 
Repo: https://github.com/ZechProjects/GeminiCraft
Demo: https://gemini-craft.vercel.app/

- **Team:** [Zech Ng](https://cerebralvalley.ai/u/Zech)
- **GitHub:** https://github.com/ZechProjects/GeminiCraft
- **Demo video:** https://youtu.be/DzNl-qwa_tA
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=7

### 80. AXL

Automated, design-driven software generation. Axl bridges Google Stitch (UI design tool) with a self-improving agentic development loop — producing and continuously refining working software from a visual specification.

- **Team:** [Florian Kowarsch](https://cerebralvalley.ai/u/FloKo), [Ian Yip](https://cerebralvalley.ai/u/ianyip), [NIcholas Sadjoli](https://cerebralvalley.ai/u/S_SSG), [Volo IER](https://cerebralvalley.ai/u/Myr)
- **GitHub:** https://github.com/ajentik/axl
- **Demo video:** https://drive.google.com/drive/folders/1JJ4obFNiXIAvxR-OkQYYNz-7EIgrkkZN?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=8

### 81. srikaya jam

NOVA (Navigable Open-World Virtual Agent) is an autonomous agent built on Mistral that can reason about goals, navigate complex 3D environments, and execute multi-step plans in dynamic virtual worlds. Instead of mapping instructions directly to actions, NOVA converts structured environment states into textual representations and uses a Mistral-based planning module to generate hierarchical action sequences. This project demonstrates how general-purpose language models like Mistral can serve as structured decision-making engines for embodied AI systems, with applications extending beyond games to robotics, simulation, and autonomous agents operating in real-world environments.

- **Team:** [King Binjai](https://cerebralvalley.ai/u/Dmark003)
- **GitHub:** https://github.com/Danielmark001/mistral
- **Demo video:** https://www.youtube.com/watch?v=UAuYM3YuMEM
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=9

### 82. Thought Experiment

Principia is a multi-agent pipeline built on Gemini that translates natural language into mathematically rigorous, visually stunning numerical simulations. To overcome LLM hallucinations, Principia introduces a "System 2" workflow with a self-correcting loop. First, a Formalizer agent extracts strict math (ODEs, PDEs, SDEs, Quantum mechanics) from user prompts. Then, a Critic peer-reviews it, running dimensional homogeneity cross-checks using SymPy via a Python execution tool. If flawed, the Critic forces the Formalizer to revise using contextual feedback. Once verified, an Execution agent compiles the math into SciPy solvers and Matplotlib rendering scripts. Wrapped in a sleek Next.js UI inspired by Google Antigravity, Principia actively reasons about physics to bring complex concepts to life, paving the way for AI-driven scientific discoveries.

- **Team:** [Yungi Jeong](https://cerebralvalley.ai/u/yungijeong)
- **GitHub:** https://github.com/realYungi/Principia-AI?tab=readme-ov-file
- **Demo video:** https://youtu.be/wa9WcIpORok
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=10

### 83. ScamBreak

Global scam operations stole >$1 trillion last year, with video call fraud surging 1,500% across APAC. Scammers run organized call centers with fake uniforms, staged offices, and counterfeit badges, video-calling victims to threaten arrest in real time. Yet every anti-scam tool works after the damage is done. Nothing protects victims during the call. 

ScamBreak changes that. Four Gemini agent swarms work in parallel to analyze live video calls in real time: scanning for fake badges and staged backdrops, detecting psychological manipulation, validating against government protocols, and delivering a verdict within seconds. 

When a scam is detected, it intervenes with clear warnings designed for victims under stress. With one click, it auto-generates a complete incident report with AI evidence, screenshots, transcript analysis, and visual identity search, ready to submit to authorities. ScamBreak doesn't detect scams after they happen. It breaks them while they're happening.

- **Team:** [Uncanny Gold Owl](https://cerebralvalley.ai/u/uncannygoldowl)
- **GitHub:** https://github.com/hanselhansel/scam-protector
- **Demo video:** https://youtu.be/P3Fn1cUdNCA
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=11

### 84. visionmusic

It turns any camera into a live music instrument. Point it at something and  plays music that matches what it sees — dark room gets ambient drone, bright party gets upbeat pop, nature gets acoustic folk.  Generate a creative experience: bridging sight and sound in real-time. Filler music to prevent awkward silence. Like giving someone music on demand. You don't pick the music — the world around you does.

- **Team:** [tk tan](https://cerebralvalley.ai/u/tantk)
- **GitHub:** https://github.com/tantk/visionmusic
- **Demo video:** https://www.youtube.com/shorts/cyr2drfo7IA
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=12

### 85. Li Yang

Agent skill wrapper over my existing** Qwen3TTS single-speaker finetuning script translated from PyTorch to MLX.

High quality TTS for any Apple Silicon Mac with >16GB RAM. Can hook up to OpenClaw's on such devices to respond to you in a voice you like.

- **Team:** [Li Yang Hew](https://cerebralvalley.ai/u/hewliyang)
- **GitHub:** https://github.com/hewliyang/qwen-tts
- **Demo video:** https://www.youtube.com/watch?v=jdZdwBvpZNU
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=13

### 86. Purrfect Match

Purrfect Match is a mobile app that transforms the experience of finding a cat to adopt or buy. Instead of going into centres with no idea of what to expect, users swipe through cat profiles posted by verified stores and adoption centres, match with cats they love, and chat directly with owners to arrange everything in one place. It also features a community feed where cat lovers can share photos, stories, and advice, making it more than just a marketplace but a dedicated social space for cat lovers. In today's day and age, people have become so accustomed to this boring and inefficient system of pet adoption that there have been no improvements or changes. Purrfect Match not only improves the quality of life of both cat lovers and cat shop owners, but also creates a brand new engaging experience of cat adoption

- **Team:** [Kiaan Shah](https://cerebralvalley.ai/u/ks123)
- **GitHub:** https://github.com/KiaanBest/Purrfect-Match
- **Demo video:** https://youtu.be/q531T3nsAi4
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=14

### 87. Singularity

Rendara is a fully autonomous AI video editing system that acts as a personal content team. Driven by Google Gemini and Google Cloud STT, Rendara allows users to edit videos simply by chatting naturally. The AI intelligently handles rough cuts, silence removal, and context-aware B-roll insertion, transforming raw footage into polished content without a traditional timeline.

- **Team:** [Umang Gupta](https://cerebralvalley.ai/u/umangguptacs)
- **GitHub:** https://github.com/Usgupta/rendara
- **Demo video:** https://youtu.be/xplBvJ53Jls
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=21

### 88. sneha

Unbound is a music genre discovery app that breaks listeners out of their Spotify suggestion bubbles by mathematically finding the music they are wired to love but have never discovered. Instead of relying on AI to hallucinate recommendations, Unbound fetches real listener tags from Last.fm, maps the user's "taste fingerprint," and uses a K-Nearest Neighbors (KNN) algorithm with cosine similarity against 40 global genre centroids to pick the top 5 unexplored matches. Gemini 2.0 Flash is then used purely for the explanation layer, translating the mathematical proximity into beautifully written, emotionally resonant reasons for why the genre fits the user's sound identity. It is the personalized, fully computed successor to the famous app - Every Noise at Once.

- **Team:** [Sneha B](https://cerebralvalley.ai/u/snehab_74)
- **GitHub:** https://github.com/Sneha421/unbound
- **Demo video:** https://drive.google.com/file/d/1W8-8L8R2eZZd82bkpWn7uVz7uEkkGqN-/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=30

### 89. AltoGen

ALTO is an AI-powered preview tool that shows what the view could look like from a specific location and height.

Users enter an address, floor level, viewing direction, and time of day. ALTO uses Google Maps, Places, and Geocoding to lock onto the exact spot, then pulls real context from Street View, satellite maps, and Elevation data. The Street View images are fed directly into Nano-Banana as reference images, so the generated skyline stays tied to the real location. Gemini writes the short description, and Nano-Banana generates the final panoramic preview image.

This solves a common problem: people cannot easily judge a property’s actual view remotely or when it hasn't been built yet. ALTO helps curious users and potential buyers/renters compare options faster and make better decisions with less guesswork.

- **Team:** [Andrew Yu](https://cerebralvalley.ai/u/andnyu)
- **GitHub:** https://github.com/andrewnyu/alto
- **Demo video:** https://youtu.be/PNt62QFlz-g
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=31

### 90. Polysignal

PolySignal is a forensic intelligence terminal for prediction markets. It detects information asymmetry by cross-referencing on-chain trades with global news. Specifically, our engine flags volume anomalies and calculates the temporal gap between trades and headlines. We generate alpha scores to unmask the insiders and provide it to the ordinary punters.

- **Team:** [Kevin Y](https://cerebralvalley.ai/u/kayeezy), [Amitansh Gupta](https://cerebralvalley.ai/u/amitanshg)
- **GitHub:** https://github.com/kayyfl/polysignal
- **Demo video:** https://youtu.be/HkU5mN12QN8
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=42

### 91. Pixel2Pieces

Pixel2Pieces is a first-of-its-kind "Generative Manufacturing" pipeline that transforms 2D images into physical 3D puzzles. While existing AI models focus on generating digital meshes for games or virtual reality, Pixel2Pieces empowers users to turn a single photo—be it a pet, a landmark, or a personal sketch—into a tangible, interlocking sculpture.

By leveraging the multimodal reasoning of Gemini 3 Flash, our application analyzes the spatial volume of an image and automatically generates a series of 2D flatpack blueprints (SVGs/PDFs) with precise interlocking notches. These can be printed on common cardstock, laser-cut, or CNC-routed for immediate manual assembly.

- **Team:** [Tracy Shen](https://cerebralvalley.ai/u/arctyf), [Yan Wen Teng](https://cerebralvalley.ai/u/yanwen_t)
- **GitHub:** https://github.com/alemonmelon/Pixel2Pieces
- **Demo video:** https://drive.google.com/file/d/1kKB1g7winxgf4i1llkTUtb-f79CLCnCx/view?usp=drivesdk
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=15

### 92. OptikOps

OptikOps is the first Agentic Detection & Response (ADR) platform powered by Gemini, built to protect websites from AI agents, not just humans.

Traditional security assumes attackers are human. In 2025 that changed. Autonomous AI agents now scrape pricing, map APIs, reconstruct business logic from JS bundles, and extract proprietary knowledge while appearing as legitimate traffic. WAFs and rate limiters cannot detect them.

OptikOps scans any URL and performs semantic security analysis across MITRE ATLAS, OWASP LLM Top 10 (2025), and OWASP Top 10 (2025). Gemini analyzes websites like a hostile AI agent would and returns structured threat reports with severity, affected routes, and a 0–100 MoatScore™.

Traditional scanners detect syntax. Gemini detects intent.

Live Demo
https://optikops.vercel.app

- **Team:** [Jun Min Ho](https://cerebralvalley.ai/u/jimmyho)
- **GitHub:** https://github.com/rolelify/optikops
- **Demo video:** https://www.youtube.com/watch?v=AFbEQGwTq9M
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=32

### 93. Pre-women's day Happy!

Oyako is a Gemini 3-powered platform that transforms travel photos and descriptions into personalized manga-style journals.

AI Storytelling: Uses photo analysis and user input to turn fragmented memories into cohesive, emotional narratives.
Triple Personalization: Combines user details, visual context, and AI reasoning to create unique character personas and scenes.
Interactive Features: Offers panel-by-panel editing and smart route mapping for easy social media sharing.

- **Team:** [Kitty Wang](https://cerebralvalley.ai/u/helloKittyCat), [Angie Ng](https://cerebralvalley.ai/u/Angie)
- **GitHub:** https://github.com/kokhwangng/oyako_travel_aistudio_hackathon
- **Demo video:** https://youtu.be/OFhJc4O6D7c
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=16

### 94. Codex Warriors

Type any task, and we'll auto-generate a benchmark, run it blind across frontier AI models, then have Gemini write harder follow-up tests from the failures and rerun them so you don't just see who scored best, you see exactly where each model breaks.

- **Team:** [Brian Chew](https://cerebralvalley.ai/u/brianchew)
- **GitHub:** https://github.com/bchewy/gemini-hackathon-7march
- **Demo video:** https://cap.so/s/4rqbz9jvschmwzy
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=17

### 95. Signal Path 3D

SignalPath 3D is a Nuxt 3 web application that simulates WiFi coverage in a reconstructed 3D apartment layout and recommends optimal router placement.

It uses Gemini to build the 3d layout from an input image of a floor plan

- **Team:** [PRANOY DEV](https://cerebralvalley.ai/u/devpranoy)
- **GitHub:** https://github.com/devpranoy/signal-path-3d
- **Demo video:** https://www.youtube.com/watch?v=D98Z27YmWaw
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=24

### 96. potato

Circle, a social storytelling app that turns real-life meetups into fictional episodes. Users create an Arc, add their friends, and every time they gather, they can upload a photo, a short note, or even an image from the moment together with a cut of their gathering (audio or video). The app then will convert speech to text and uses Gemini to transform that real gathering into the next episode of their ongoing story world.

- **Team:** [Fang Yee Lim](https://cerebralvalley.ai/u/fangyee)
- **GitHub:** https://github.com/TisuPaper/gemini3
- **Demo video:** https://youtu.be/J1aVTpo75m8
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=34

### 97. ProperAI

ProperAI is an agentic property discovery tool for Singapore's property market that transforms a fragmented, regulation-heavy search across 28 postal districts into a single guided experience.

Two personas sit at the core of the product. Buyers get an intelligent way to research property areas against their personal criteria, surfacing district recommendations that account for budget, commute, school proximity, lifestyle fit, and Singapore-specific eligibility constraints in one place. Property agents get a credible research tool to quickly narrow down what each buyer actually needs, track multiple lead interests and qualifications, and walk into client conversations with data-backed district shortlists rather than guesswork.

- **Team:** [Daryl Lee](https://cerebralvalley.ai/u/darlee), [Tommy Tien](https://cerebralvalley.ai/u/tommytien), [Anthony Nguyen](https://cerebralvalley.ai/u/Anthony889)
- **GitHub:** https://github.com/dleekh/properai-ghack-7Mar
- **Demo video:** https://www.loom.com/share/f275d2b4bb40456c85bfcc3aedfaaa36
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=18

### 98. Self-evolving agentic architecture

This project is based on a self-evolving agentic architecture, which is led by an architect, and acts on feedback received from a feedback agent (in this case, manifesting Warren Buffet). The objective is to identify an optimal agentic architecture for stock valuation automatically. This might save AI engineers a lot of time doing manual research and trying different architectures.

- **Team:** [Nathan Toh](https://cerebralvalley.ai/u/nathan_toh)
- **GitHub:** https://github.com/ziyu-toh/cv-gemini-hackathon-2026
- **Demo video:** https://youtu.be/xehgM8VoIEg
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=20

### 99. HackSignal

A database tool for hackathon participants seeking inspiration on how winning teams use sponsor APIs. Features a sidebar listing hackathons (discovered via web search agents), with drill-down views showing teams per hackathon -- winners are highlighted. Insight agents analyze overall sponsor API usage patterns across each hackathon and surface interesting projects. Each project includes an API analysis agent that describes how sponsor APIs are specifically used within that team's submission, catching non-obvious integrations. Web search agents can find new hackathons, import team project description URLs, or accept pasted descriptions for a summary agent to generate project cards. All agent-proposed updates go through human-in-the-loop approval, letting users accept or decline changes before they're applied.

- **Team:** [Alice Gwee](https://cerebralvalley.ai/u/alicegjw)
- **GitHub:** https://github.com/alice-gjw/HackSignal
- **Demo video:** https://drive.google.com/file/d/1qyMwWbE-O6m9HCaPYk3baBDXkxIKr-1g/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=22

### 100. GooglyEyes

I have 3 young kids in the Singapore school system.  I wanted to build a tool for them to improve their learning and knowledge retention.  One of the best ways to do this is repetition.  But repetition is not fun.  So I wanted to build up a large database of questions, digitize it and then deliver it in a form that's easy for them to consume and want to use.  Future enhancements include simple gamification, more and brighter colors, positive feedback loops, LLM support when they have questions.  but everything is built around having a large database of questions and answers to work with.  The questions are from many different PDF sources so made use of a conbination of Google Antigravity, Google Vertex, Firebase, Google Cloud and Gemini to start learning the data and manipulating into a database with question/answer pairs and different types of question models

- **Team:** [Siva Valliappan](https://cerebralvalley.ai/u/SivaInSG)
- **GitHub:** https://github.com/siva73-git/eduquiz-ai
- **Demo video:** https://youtu.be/XJWsT0VCgtE
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=23

### 101. NewsPlay

NewsPlay is an innovative AI-driven platform built for the Google Gemini Hackathon that transforms the world's static news cycle into an interactive, educational, and fun gaming experience. Instead of just passively consuming headlines, users can engage with the news through dynamically generated game mechanics.

- **Team:** [Kyrie Bai](https://cerebralvalley.ai/u/xxxy), [Hayden Red](https://cerebralvalley.ai/u/Hayden_Red), [Mark Yang](https://cerebralvalley.ai/u/paisleypark2048), [Junda Zhang](https://cerebralvalley.ai/u/purelove)
- **GitHub:** https://github.com/AHEADer/FlexGameFactory
- **Demo video:** https://drive.google.com/file/d/1k9_gLHTFYAH65j0bF-PxqNsvh0VXRE7v/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=36

### 102. OpenAnts

While a human team is still drafting their first coordination email, OpenAnts has already resolved the entire supply chain consensus in under 10 seconds.

OpenAnts is a high-speed orchestration protocol powering the next generation of autonomous machine commerce. Dozens of specialized agents—suppliers, logistics providers, insurers, and compliance officers—can discover, bid, and reach a binding consensus in seconds. On top of this foundation, we’ve built OpenAnts Platform, a ready-to-use suite that enables businesses to orchestrate multi-party operations seamlessly. It’s the Digital Nervous System for the Autonomous Swarm Era, moving global trade from slow, manual processes to near-instant, coordinated action.

- **Team:** [huoru liu](https://cerebralvalley.ai/u/huoru), [Cheng Tay](https://cerebralvalley.ai/u/ChengTay), [John Jee](https://cerebralvalley.ai/u/Pinkgu1024)
- **GitHub:** https://github.com/OpenAntNest/OpenAnts
- **Demo video:** https://youtu.be/1zc0GcduIQ4
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=48

### 103. baarse

NAIRI, Native Android Inspection & Risk Intelligence

NAIRI is an AI-driven, orchestrated mobile application vulnerability analysis tool. It streamlines the process of reverse engineering Android APKs by leveraging large language models (Google Gemini) alongside specialized static and runtime analysis environments.

- **Team:** [Artavazd Balaian](https://cerebralvalley.ai/u/Aet)
- **GitHub:** https://github.com/REASY/nairi
- **Demo video:** https://youtu.be/1-aQg606K5I
- **Project:** https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery?project=69

---

Markdown version of https://cerebralvalley.ai/e/gemini-3-singapore-hackathon/hackathon/gallery. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt