LANVAR
← The shelf

NURO

Did its thing

A wearable memory aid that records life as it happens and answers “who did I meet today?” with semantic search over your own conversations.

✳ 6th hackathon win PythonMongoDB AtlasOpenAISolace PubSub+ 2024
How it actually works
who did I meet today? who did I meet today? “you met Sam at the pharmacy.” FROM 2 RETRIEVED MEMORIES “you met Sam at the pharmacy.” FROM 2 RETRIEVED MEMORIES WEARABLE records the day as it happens EMBED ON INSERT MONGODB · VECTOR MEMORY LIVE BUS THE CAREGIVER SIDE caregiver dashboard geofences · reminders check-ins, in real time LIVE ask, and the day answers back
FIG 1 · the memory loop, animated — record, embed, ask, sweep, answer
wearable records life as it happens speaker diarization who said what MongoDB Atlas conversation store OpenAI embeddings triggered on every insert — the memory organizes itself memory lapse “who did I meet today?” semantic search over the embeddings ChatGPT + retrieved context grounded, not guessed a real answer independence, restored Solace PubSub+ live event bus caregiver dashboard geofences · reminders · check-ins live stream off the device ↴
FIG 2 · the machinery — record → embed → search → answer, caregivers on the live bus
The story

NURO is the second life of an old idea (see LIA, its 2021 ancestor) — rebuilt with Rehean Thillai, Joerex Thambaiah, and Abilash Sasitharan once ChatGPT and vector databases made the original vision actually buildable.

The wearable records real conversations and distinguishes who is speaking. Every conversation lands in MongoDB Atlas, where an insert trigger runs it through OpenAI embeddings — so the database organizes itself as memories accumulate. When a memory lapse hits, a semantic search pulls the relevant conversations and hands ChatGPT tailored context to answer with, instead of a hallucinated guess.

Solace PubSub+ streams live data between the device and a caregiver dashboard: location boundaries, reminders, remote check-ins. The eleventh hackathon, the sixth win, and the rare hack that a family member could actually want.