# DeckTwin.com

- **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:** [Areeb pasha](https://cerebralvalley.ai/u/areebp)
- **GitHub:** https://github.com/areeb1501/deck-twin-library
- **Demo video:** https://youtu.be/DmBhJowUwpE
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2926

Deck Twin is a 3D digital twin of an offshore oil and gas field that runs in the browser. It models five platforms linked by bridges, with nearly 100 pieces of equipment, 155 process lines and 360 instruments. The whole plant is built from a plain data model and driven by a live process simulation.

The problem is that offshore plants are hard to see. Operators work from flat control-room screens, safety teams watch CCTV by hand, and new staff learn the plant from drawings. When something goes wrong, the facts are spread across many systems.

Deck Twin brings them together. You can walk the decks in first person, click any piece of equipment to see its live values, and follow oil and gas from the wells to the export risers. A control-room console works like a real one, with process pages, alarms, shutdowns, fire and gas, maintenance, spares and costs.

Safety is the heart of it. Cameras around the plant send frames to Meta's SAM 3.1, which finds people, hard hats, vests, flames and smoke. Our rules turn what it finds into alerts: someone without a permit walking into a hazardous zone, a worker without a hard hat, or a fire. Operators can type what SAM should look for, and replay the last 10 seconds of any camera.

Muse Spark is the assistant. It reads only what an operator can see, explains what is happening, and suggests the next step. Nothing happens until the operator confirms it, and the plant is checked again afterwards. Operators can also describe a training drill in plain words, and Muse Spark turns it into a scenario the simulator checks and runs.

We also built a library of standard 3D plant parts (vessels, pumps, valves, piping) with a command-line tool and an AI skill, so anyone can turn their own drawings into a new plant.

The plant data and readings are simulated, and the app says so. The engineering follows real industry standards, but it is not a real asset.

## More from Meta Global AI Developer Hackathon

- [Code_WithAI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2923)
- [bro_dev](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2924)
- [Hawkaii](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2925)
- [Hallucinated Dev](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2927)
- [Sightloop](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2928)
- [Abu Bakar Siddik](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2929)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2926. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt