What we've built, and what's next.
Every session ends with running code, not slide decks. Here is the verifiable record of our first 3 in-person build sessions in Jaipur.
64 developers deployed autonomous agent loops from scratch without relying on black-box abstractions. We wired local vector databases directly into custom tool-dispatchers, tested failure recoveries in real time, and benchmarked memory footprints across heterogeneous student laptops.
We took raw open weights and quantized them down to 4-bit integer representations using llama.cpp and AWQ. Every participant set up local terminal inference on their own hardware with zero cloud API tokens or external rate limits. 38 custom inference endpoints were verified live during the session.
Our inaugural session in Jaipur stripped away hype to write pure execution loops from first principles. Builders implemented token streaming, state persistence, and self-correcting validation handlers in vanilla Python. Everyone left with a working local CLI agent that could run system diagnostics autonomously.
Seats are limited by the physical capacity of the room. RSVP early to claim a build slot.
Developers with working Python fundamentals who want to run autonomous models without paying recurring cloud token bills.
A compiled, running SLM instance running entirely on your local machine with customized hardware-caching logic.
Can't make it to this session in person?
Check out our curriculum breakdown on /learn or talk to active builders between events in the Jaipur Community Hub on /community.
