The Wharton School - AI startup founder & Wharton MBA student★ FinalistDeep-reviewed
by MANAS SHARMA
What they're building
At this week's Google I/O, Demis Hassabis highlighted DeepMind's focus on World Models as a critical step toward AGI. With real-time video editing via natural language already here, it got me thinking about the immediate applications for entirely new, dynamic user interfaces. What if we could replace crowded, busy fitting rooms in retail stores with intelligent screens? After 4 hours of rapid prototyping and letting Antigravity entertain my ever-changing requirements, we built "Live Look!" It's a real-time virtual try-on app that completely changes the browsing experience: - It uses the Lucy VTON model (powered by Decart) to render garments in real-time. - It also lets users browse the product catalog, swipe, and "add to cart" entirely through hand gestures (powered by Google's MediaPipe).
AI code reviewrepo: no-repo
No repository provided (repo_url empty); demo_url is a LinkedIn post and site_url is a LinkedIn profile, so there is no inspectable code and no live product to verify - treated as no-build per rubric. The stated stack is Decart's Lucy VTON model plus Google MediaPipe for hand-gesture browsing - none of the three hackathon sponsors (Composio, Nebius, Tavily) are used. Described as a 4-hour real-time virtual try-on prototype; concept is interesting but unverifiable and it is not an AI-agent build. ⚑ No repo and no verifiable live demo (floors applied). Zero sponsor-stack usage. Not an agent. Only evidence is a LinkedIn showcase post.