Full Stack / 2026 
The problem
Creating listings for a garage sale means identifying and describing several items from the same scene.
- RoleDeveloper · team of five
- ContextOpenAI Buildathon Manila
- ResultTop 3 at OpenAI Buildathon Manila
What we built
On a five-person team, I built a mobile-first marketplace that turns a scene photo into editable product listings and a live, shareable garage sale. Sellers review the detected items, edit the details, and confirm prices before publishing.
BUILT WITHOpenAI VisionSAM 2Convex
Technical decisions
Seller review before publication
The flow includes item inclusion and exclusion controls, editable metadata, and seller-confirmed pricing.
Fallbacks for image segmentation
Progressive detection uses SAM 2 segmentation with a bounding-box fallback so sellers can continue reviewing items when segmentation is incomplete.
Results
- Placement
- 2nd Runner-Up
- Participants
- 250+
- Placed 2nd Runner-Up among 250+ participants at OpenAI Buildathon Manila.
- Built the photo-to-listing flow on a five-person team.
