Congratulations to the winners of the Persistent Context Sprint Hackathon at #MongoDBlocal Build Fest!
Builders had a few hours on-site to ship AI-native apps where memory, state, context, and retrieval all live on MongoDB, no external vector store or cache required.
🥇1st Place: Amelia – Brendan Giang, David Wu, Jerry Yan, and Boris N. (Team Amelia). Amelia is a live captioning assistant for deaf and hard-of-hearing users that separates speakers into voiceprints and stores each person's identity, conversation history, and commitments directly in MongoDB Atlas, using vector search to retrieve relevant context in real time so every conversation picks up right where it left off..
🥈 𝟮𝗻𝗱 𝗣𝗹𝗮𝗰𝗲: TraceCase – Jiyun K. and Carl Okpala. TraceCase turns a vague bug report into a reproducible case, asking for the missing details, then recreating the bug across browsers, operating systems, and devices in isolated environments, with MongoDB storing all case context and evidence. If TraceCase finds and verifies a fix, it prepares the GitHub PR itself.
🥉 𝟯𝗿𝗱 𝗣𝗹𝗮𝗰𝗲: Perpetual – Rohan Gandotra and Sharique Khatri. Perpetual logs every agent session to MongoDB Atlas, mines those logs for repeated routines with aggregation pipelines, and compiles them into new skills stored as a vector search collection. A change stream pushes each new skill to every other connected agent instantly, so one agent's experience becomes the whole team's ability.
Thank you to everyone who built and demoed, and to our partners, Cerebral Valley, for helping make the hackathon possible.
