# TechLabID

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Hie Cua Als Diman Susanto](https://cerebralvalley.ai/u/TechLabID)
- **GitHub:** https://github.com/TechLabID/AIRA-Companion
- **Demo video:** https://www.youtube.com/watch?v=l33VOqxX7r8
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/713

**AIRA Companion — Bringing AI Beyond the Screen**

Most AI assistants are limited to conversations. They can answer questions and generate suggestions, but they cannot reliably connect their understanding of the physical world to safe, verifiable actions. This creates a critical challenge for embodied AI: how can an AI assistant perceive its environment and assist people without compromising safety, privacy, or human control?

AIRA Companion addresses this challenge by combining Meta AI technologies, computer vision, voice interaction, and an independent safety verification layer with a Unity-based simulated environment.

The prototype integrates Meta Muse for voice and intent processing, Meta SAM 3.1 for visual perception, and local YOLOX-S detection. Instead of allowing AI-generated decisions to trigger actions directly, AIRA applies independent checks for user consent, evidence freshness, target validity, and safety. Only authorized actions can proceed to bounded navigation in Unity; unsafe or insufficiently verified requests are deferred.

Our prototype demonstrates voice interaction, visual awareness, safety-aware decision-making, and simulated navigation, including arrival and return-home behaviors. Real Meta model inference and local perception have been tested, while physical autonomous drone flight remains outside the current verified scope.

AIRA's innovation is not simply making AI see or respond—it is establishing a trustworthy connection between AI perception and action.

Our long-term vision is to make embodied AI companions more accessible, privacy-aware, and useful for assisted living, education, and everyday human support.

## More from Meta Global AI Developer Hackathon

- [NexussAI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/710)
- [Ratnesh Saxena](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/711)
- [Muse Studio Lite](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/712)
- [Pawception](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/714)
- [alloy-agent](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/715)
- [Lumina](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/716)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/713. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt