# MoveIn - Columbia

- **Event:** [Gemini 4 NYC Hackathon](https://cerebralvalley.ai/e/gemini-4-nyc)
- **When:** Sat, Nov 21 at 9:00 AM – 10:00 PM (EST)
- **Where:** New York, NY
- **Team:** [Uddhav Narayan Gilda](https://cerebralvalley.ai/u/ung5), [Nakul Ramesh Limbani](https://cerebralvalley.ai/u/nakullimbani), [Malhar Paresh Mangtani](https://cerebralvalley.ai/u/MalharMangtani)
- **GitHub:** https://github.com/Uddhav786/movein
- **Demo video:** https://www.youtube.com/watch?v=K7uEfGgnj4g&feature=youtu.be
- **Gallery:** https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/107

MoveIn: film your room, get it furnished with real products that actually fit.

Renters in small apartments buy furniture that looks right online but blocks a closet door, covers a radiator, or doesn’t leave room to walk. MoveIn starts from one narrated phone walkthrough. Gemini maps the room: walls, windows, radiator, closet, every door and its swing, the furniture that’s staying or leaving, and what the renter said (“I need a gaming setup, cozy and warm”).

The Decisions API asks only for the missing evidence that would change a decision (“measure the room”, “show that door opening”). A tiny on-device metric depth model measures what’s fully in frame.

Gemini 4 Carbon plans what the room needs. Shop MCP finds live Shopify products, and Gemini 4 Neon reads each listing’s assembled size with a quote from the listing.

FitProof, plain deterministic code with no AI, places each piece against a wall, clear of door swings, windows, the radiator, existing furniture and walkways, or says “don’t buy this” with the reason in inches. Anything only estimated from video gets a safety buffer, and estimates alone never produce a “fits”.

The renter sees a to-scale 3D room and downloadable 2D/3D layouts. They also get a photoreal version of their own video with the chosen product placed in it: Nano Banana edits a key frame, OpenCV cuts out only the new piece, follows the camera and places it into every frame, so the rest stays real footage. Buy does a live re-check of the exact variant and opens the merchant’s own checkout.

Evaluation: FitProof has 0 false fits on our benchmark versus 2 for the “is it narrower than the wall?” baseline, and the test suite passes.

## More from Gemini 4 NYC Hackathon

- [Capture Studio](https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/104)
- [GitHappens](https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/105)
- [StageState](https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/106)
- [CodeWarrior](https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/108)
- [DocentAI](https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/109)
- [Flare](https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/110)

---

Markdown version of https://cerebralvalley.ai/e/gemini-4-nyc/hackathon/gallery/107. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt
