# Global Codex Hackathon: [Singapore]: Project Gallery

- **Event:** [Global Codex Hackathon: [Singapore]](https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual)
- **When:** Tue, Apr 21 at 9:00 AM – 9:00 PM (GMT+8)
- **Where:** Online
- **Hosts:** [Cerebral Valley](https://cerebralvalley.ai/u/cv), [OpenAI](https://cerebralvalley.ai/u/openai)
- **Projects:** 21 (3 placed)
- **Page:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery

## Projects

### 1. Pathfindr

SkillForge is a local-first system that helps Codex learn a team’s repo-specific conventions over time. When developers correct Codex in the same way more than once, SkillForge turns that repeated correction into a candidate SKILL.md, evaluates whether that skill actually improves performance, and only then ships it into the active workspace. A cloud control plane stores findings and powers a dashboard so the team can inspect what was learned and why.
The dashboard url - https://dashboard-ten-ruby-38.vercel.app/

- **Team:** [Shubham Pai](https://cerebralvalley.ai/u/shubhamm), [Ashard Nizar](https://cerebralvalley.ai/u/anzr), [Tim Gaul](https://cerebralvalley.ai/u/fourixai)
- **Demo video:** https://preview.screen.studio/share/X4QviS8t
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=16

### 2. ResolvAI

Project Description

This project is a self-healing production system powered by Codex that automates the full lifecycle of incident resolution. Instead of relying on engineers to manually investigate and fix failures, the system connects existing signals—logs, stack traces, code, and tests—into an intelligent workflow that can detect issues, diagnose root causes, and generate fixes automatically.

It introduces a structured, human-in-the-loop approach where incidents trigger automated analysis, and engineers step in only for validation and control. The result is a faster, more reliable, and scalable way to handle production failures, reducing downtime and operational cost.

Key Features
1. Automated Incident Detection & Triggering
Production failures automatically generate structured incident artifacts (logs, stack traces, context).
A watcher system detects these incidents and triggers the resolution workflow instantly.
2. AI-Powered Root Cause Diagnosis
Uses Codex with an incident-diagnoser skill to:
Analyze logs and stack traces
Inspect the codebase
Identify the root cause
Produces a diagnosis report without modifying code.
3. Human-in-the-Loop Checkpoint
Engineers review the diagnosis before any changes are made.
Ensures control, trust, and safety in automated workflows.
4. Guided Automated Remediation
Codex uses an incident-remediator skill to:
Fix the bug
Add regression tests
Create a branch and commit changes
Engineers can refine or guide the process through the IDE.
5. End-to-End Workflow Orchestration
Covers the full lifecycle:
Detection → Diagnosis → Review → Fix → Delivery
Moves beyond code generation to process automation.
6. Extensible Delivery Pipeline
Supports future integration with:
Automated pull request creation
CI/CD validation pipelines
Safe deployment gates
7. Scalable Multi-Service Architecture
Can expand beyond a single repo to:
Multiple services and repositories
A knowledge graph capturing dependencies and lineage
Enables ecosystem-wide issue resolution.
8. Real Business Impact
Faster incident resolution (minutes instead of hours)
Up to:
70% reduction in manual investigation
36% lower downtime-related losses
15–35% reduction in IT operations costs

- **Team:** [Arushi Tejpal](https://cerebralvalley.ai/u/rushi_T), [Nivedita Kumar](https://cerebralvalley.ai/u/nivk)
- **Demo video:** https://drive.google.com/file/d/1aLre_bMaWH46CQOWUi2jpQSUJ-DP3hTo/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=17

### 3. AgentBridge Legends ( Mantel )

Universal Agent Context Bridge is a prototype that evaluates how suitable an API is for AI agents, explains the risks, selects a safer subset of endpoints, and generates an MCP starter bridge. It combines deterministic scoring, safety-oriented curation, and report generation to help teams turn raw API specs into more agent-ready tooling.

- **Team:** [James Moyle](https://cerebralvalley.ai/u/pi3ce), [Nicole Ramirez](https://cerebralvalley.ai/u/nxrmrz), [Rico Wismans](https://cerebralvalley.ai/u/rico_mantel)
- **Demo video:** https://youtu.be/vwUarA3R7fI
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=20

### 4. Genie Capsule

The original application was developed using Java 1.5 with Servlets and JSP, a legacy technology stack that was common at the time. While functionally sound, the experience around it began to show its age. The UI was not intuitive, troubleshooting required deep technical knowledge, and users had to dig manually through screens, logs, and data to understand what went wrong when issues occurred.
Users spent too much time reacting to issues instead of proactively managing them. Tracing errors, understanding exceptions, and debugging problems required effort that slowed operations and created friction across teams.
That’s when we made a strategic decision:
don’t just upgrade the technology—reimagine the experience

- **Team:** [Renuchandran Kasi](https://cerebralvalley.ai/u/Renuchandran), [Ranjith Ratna Kumar Subramanian](https://cerebralvalley.ai/u/Ranjith12)
- **Demo video:** https://youtu.be/FrFvrbf3krE?feature=shared
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=21

### 5. Time Under Tension

Launchpad is a Codex-powered starter platform for rapidly turning a client brief into a working prototype without leaving the Codex window. It combines reusable software modules, custom skills, requirements generation, prototype implementation, and GitHub automation into a single workflow that helps our team go from idea to shipped software much faster.

- **Team:** [Jackson Gayda](https://cerebralvalley.ai/u/jgtut), [Dave Gill](https://cerebralvalley.ai/u/Daveg)
- **Demo video:** https://drive.google.com/file/d/1MnYZvW85XNcz3zQQAQozNN3xyrd7mRcK/view?usp=drive_link
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=1

### 6. AutoCTO

AutoCTO analyses GitHub organisations using a multi-agent Codex pipeline. It reviews repos for code quality, architecture, and tech debt, then surfaces findings through a dashboard and interactive knowledge graph.

- CTO Dashboard - Org-wide view of repo health, high-risk issues (P0–P3), improvement PRs, and recommendations. Built with Codex.
- Multi-Agent Pipeline - 10+ Codex subagents run in parallel per repo covering structure, standards, quality, documentation, and design patterns.
- Knowledge Graph - Force-directed graph showing repos, contributors, and technologies with AI chat for querying any node.
- GitHub Actions - Automated pipeline: clone, analyse, triage, report. Hands-free.
- Org-Configurable - Plug in any GitHub org. We demo'd on OpenAI's public org.

- **Team:** [Katie Danko](https://cerebralvalley.ai/u/ktomoro), [Malan Evans](https://cerebralvalley.ai/u/malanevans), [Jared Chung](https://cerebralvalley.ai/u/jaredchung)
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=2

### 7. Insurgence Worldview

Project Worldview is an AI-powered command center that transforms fragmented project and system signals into a unified, real-time operational picture.

Instead of relying on static dashboards or manually compiled updates, the platform continuously ingests data from project artifacts, system telemetry, and operational signals to surface what actually matters. It highlights emerging risks, uncovers hidden dependencies, and provides a clear, dynamic view of overall health.

At its core, Project Worldview acts as an intelligent layer over existing tools—interpreting signals rather than just displaying them. Users can select any project or service and instantly understand its true status, supported by AI-generated insights, explanations, and recommended actions.

The system is designed not only to detect issues early but to guide response. It can proactively suggest next steps and generate key artifacts such as executive updates, recovery plans, and incident summaries, reducing manual effort and improving decision speed.

By combining situational awareness with AI-driven reasoning, Project Worldview enables teams to move from reactive reporting to proactive command—making complex delivery and operational environments easier to understand, manage, and optimize in real time.

- **Team:** [Rusty Blewitt](https://cerebralvalley.ai/u/rusty), [Chris Nones](https://cerebralvalley.ai/u/CNones), [Scott Meads](https://cerebralvalley.ai/u/ScottM)
- **Demo video:** https://cap.so/s/3fvstw5bcvn35tr
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=3

### 8. Quantium Digital Twin

(+ Mario Attard)
Digital Twin is a consent-aware platform that turns interviews, assessments, public footprint, and work signals into an evidence-grounded twin that can be inspected, validated, and packaged as an installable agent/plugin.

- **Team:** [Finn Gladstone](https://cerebralvalley.ai/u/finngladstone), [Mehran Mozaffari](https://cerebralvalley.ai/u/Mehran)
- **Demo video:** https://drive.google.com/file/d/1RJXjVqCj3-300kTGj2DBJkgM5ldaZ2vg/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=4

### 9. CodexJunkies

Solution: The Studio
Multi-agent coding has an attention problem: you launch Codex, then squint at spinners, logs, and diffs trying to guess who's thinking, who's stuck, and who just changed their mind. Dashboards only add more tabs.
The Studio replaces all of that with a single 3D room. Four agents at four desks. When the Architect plans, they walk to the whiteboard and sketch live. When Backend and Frontend negotiate an API, they meet there and argue in speech bubbles. Extended reasoning happens on the couch. Debugging happens out loud to a glowing rubber duck. Code streams onto monitors as it's written.
Spatial UI scales the way attention actually does.

- **Team:** [Sarthak Jain](https://cerebralvalley.ai/u/jainai), [Puneet Singh](https://cerebralvalley.ai/u/Puneetsigns), [Mokshita Vashisht](https://cerebralvalley.ai/u/MVashisht)
- **Demo video:** https://docs.google.com/document/d/193yUkEcJI90Ox6g9kUcrN1XgVEenD4_S7GVIaoFpjPQ/edit?tab=t.0
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=5

### 10. Chieftns.ai

AI Tender Assistant is a web app that helps teams decide whether to bid on a tender and then generate a first-pass response package faster. It combines a Next.js frontend with a FastAPI backend, ingests tender documents, runs a transparent multi-agent assessment across scope fit, compliance, delivery risk, win probability, and response requirements, and then turns that analysis into a requirements-led DOCX draft that teams can review and refine instead of starting from a blank page.

- **Placement:** 1st Place
- **Team:** [Alex Owens](https://cerebralvalley.ai/u/AO_Chieftns)
- **Demo video:** https://drive.google.com/file/d/1qa4ovvrkJ3rexlBeCKYeVIsGxQxp5GIl/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=6

### 11. Cloudwerx Advocado

A way to capture fragmented thoughts. We wanted to create an accessible way for people to log notes and continue their trains of thought over time. Too often, important context gets lost. Our goal is to give users a way to capture notes, continue note threads, and receive small suggested next steps so the process feels manageable rather than overwhelming.

- **Team:** [Jacqueline Sprott](https://cerebralvalley.ai/u/Sproot), [Hoa Hoang](https://cerebralvalley.ai/u/hoa), [Andrew Chaseling](https://cerebralvalley.ai/u/AndyAI)
- **Demo video:** https://youtu.be/Z8zo7IMPgCI
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=7

### 12. Corca

Agentic tools automate code generation, but the bottleneck has moved. It's no longer how to build. It's what to build, and whether it's right. Agents don't check if a request conflicts with business goals, existing specs, or product
policies. They build fast, correctly, and wrong.

Specdown addresses this with three principles. Traceability: business goals, product goals, features, acceptance criteria, tests, and code are bidirectionally linked. Change one layer, the system knows what else breaks. Consistency: if a new spec contradicts existing specs or policies, it's caught before any code is written. Executable specs: one file is the spec, the test, and the report. Run it and you get green or red. Spec and test are the same artifact, so they can never drift apart.

We built a Codex pipeline on top of this. User feedback comes in. Codex profiles who the user is, analyzes root cause against existing specs, generates a new spec verified for consistency against business goals and policies, builds code in parallel worktrees until specs pass, reviews the PR against the spec, and merges on approval. The spec is the interface between every Codex task. What gets planned becomes the build target, the review criterion, and the deploy condition.

- **Team:** [Jaeho CHUNG](https://cerebralvalley.ai/u/Jayho), [HyunMyeong Yoo](https://cerebralvalley.ai/u/Hyunu), [Jaeyoung KANG](https://cerebralvalley.ai/u/jaeyoung)
- **Demo video:** https://www.youtube.com/watch?v=nzcBt5gY-Lg
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=8

### 13. Project Daedalus - PwC

End to end workflow of skills and a custom human annotation app for expert review and skill iteration.

- **Placement:** 3rd Place
- **Team:** [Will Baker](https://cerebralvalley.ai/u/willb), [James Carolan](https://cerebralvalley.ai/u/jimmyc)
- **Demo video:** https://youtu.be/_bgQmkOnzUY
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=9

### 14. V2.AI

CodexCalls is a real-time human-in-the-loop system that enables AI coding agents to safely make high-impact decisions by calling a human when needed.

When an AI agent encounters a risky or ambiguous action (e.g. running destructive commands, pushing code, or getting stuck), CodexCalls automatically triggers a phone call. A voice agent explains the situation, captures a spoken decision (approve, deny, or instruct), and converts it into a structured outcome that safely continues or stops execution.

This creates a seamless bridge between autonomous AI agents and human oversight without slowing down workflows.

- **Placement:** 2nd Place
- **Team:** [Aneta Nikoloski](https://cerebralvalley.ai/u/aneta)
- **Demo video:** https://drive.google.com/file/d/1yjxWZwG38MgpYhjXIeoUnf3aRVFR7T8u/view
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=10

### 15. Endava Assistant

Endava Assistant is an enterprise knowledge assistant that turns fragmented internal delivery artifacts into grounded, reusable case studies and customer stories. It combines structured metadata ingestion, an MCP retrieval server, and a chat-style web app so users can find the right project evidence quickly and generate summary-first outputs with traceability and anonymization built in.

- **Team:** [Alejandro Perez](https://cerebralvalley.ai/u/Alej)
- **Demo video:** https://www.loom.com/share/8128e4d1cfa447abb95a9aa256175c9b
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=19

### 16. Aiworks

Aiworks Pat is an enterprise AI readiness visibility system that helps organizations understand whether their teams are actually prepared to adopt AI. It combines Pat, an AI interviewer, with a readiness dashboard and a Codex-powered exploration layer to turn qualitative employee feedback into decision-ready insights.

- **Team:** [Instructors Aiworks](https://cerebralvalley.ai/u/aiworks), [Shirney Huang](https://cerebralvalley.ai/u/Shirney), [Ting Chun Kuo](https://cerebralvalley.ai/u/Pot)
- **Demo video:** https://drive.google.com/drive/u/1/folders/1nQ6HpwmXfsSngOFYsMGsAMBXShuAhCfF
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=11

### 17. Project Dea

End to end workflow of skills and a custom human annotation app for expert review and skill iteration.

- **Team:** [Kayur Patel](https://cerebralvalley.ai/u/LegendaryPhoenix)
- **Demo video:** https://youtu.be/_bgQmkOnzUY
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=13

### 18. FlowLens

Flowlens is a visual-first development tool that lets developers see and edit their application's architecture as interactive draw.io diagrams, then use natural language to instruct an AI agent (OpenAI Codex) to implement changes to the actual codebase. The diagram serves as both documentation and an input surface, edit a flow, describe what you want, and the agent updates the code to match. Key features: embedded draw.io editor with semantic diagram diffing, real-time Codex agent integration via WebSocket streaming, architecture manifest that maps diagram nodes to code references, and a reset/restore system for safe experimentation.

- **Team:** [Edison Yan](https://cerebralvalley.ai/u/edisontomoro), [Sadiah Zahoor](https://cerebralvalley.ai/u/Sadiah_Z), [Dylan Bollen](https://cerebralvalley.ai/u/Dylanbollen)
- **Demo video:** https://www.loom.com/share/765cc5d0627d4fd39b5dae85cc609284
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=12

### 19. ADEN (Autonomous Defect & Experience Navigator)

ADEN is an AI agent that automatically collects, triages, and resolves customer support tickets for a global SaaS product. It reads tickets from multiple channels (Gmail, Google Sheets), classifies them by analyzing the actual product codebase and support documentation using Codex, and takes action — answering FAQs via Slack, escalating real defects to the PM, and auto-generating code fix PRs for recurring UX gaps.

Key Features
- Multi-channel CS collection: Automatically pulls tickets from Gmail and Google Sheets into Linear
- Code-aware triage: Codex reads the product's frontend/backend code and compares it against support documentation to determine whether an issue is a defect, FAQ, or UX gap — not just text classification, but actual code-level judgment
- Auto-response via Slack: FAQ tickets are matched and answered instantly; defect tickets trigger PM alerts with severity assessment
- Codex auto-fix PR: When the same UX confusion is reported 3+ times, Codex automatically writes a tooltip/explanation into the codebase, generates a test, and opens a GitHub Pull Request

- **Team:** [Joosub Byun](https://cerebralvalley.ai/u/Joosub), [Yong Rhee](https://cerebralvalley.ai/u/kanto)
- **Demo video:** https://www.youtube.com/watch?v=wwjXYHKQzxI
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=18

### 20. HCLTech India

Legacy to Modern Architecture - Modernized the legacy AWS Card Demo COBOL application to a modern technology stack using Java, Spring Boot for the backend, and React for the frontend, while preserving core business logic. Migrated DB2 and VSAM datasets to PostgreSQL.

- **Team:** [Ashutosh Singh](https://cerebralvalley.ai/u/ashutosh_si_hcl), [Arvind BAHUGUNA](https://cerebralvalley.ai/u/arvind_b), [Rohit Verma](https://cerebralvalley.ai/u/rohitverma)
- **Demo video:** https://youtu.be/kyOqxlBsMr0
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=14

### 21. Conversation is all your charts need

An AI-powered charting agent that lets anyone transform raw data into polished, interactive charts through natural conversation — no coding required.

Key capabilities include:

- Intelligent Chart Recommendation — The AI analyzes uploaded data structure and automatically suggests the most appropriate chart type (bar, line, scatter, pie, etc.) before the user even asks.
- Natural Language Editing — Users can modify colors, axis labels, chart types, legends, and data ranges through free-form conversation, making iteration as easy as typing a sentence.
- Click-to-Edit Interactivity — Users can click directly on any chart element to target and modify just that component, bridging the gap between conversational AI and direct manipulation.
- Dual-Protocol Architecture — Built on FastAPI + FastAPI-MCP, the service simultaneously exposes both HTTP REST endpoints and MCP (Model Context Protocol) interfaces, enabling seamless integration as a standalone web app or as a composable tool within larger agentic workflows.
- Non-Developer First — Every design decision prioritizes accessibility, ensuring that business analysts, marketers, and domain experts can produce production-quality visualizations without writing a single line of code.

- **Team:** [Seungwon Choi](https://cerebralvalley.ai/u/Judy_2424), [Juhyeon Kim](https://cerebralvalley.ai/u/JooK), [yoonseok yeom](https://cerebralvalley.ai/u/yuemyoun)
- **Demo video:** https://youtu.be/WzbBXBHt_SE
- **Project:** https://cerebralvalley.ai/e/open-ai-hackathon-singapore-virtual/hackathon/gallery?project=15

---

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