AI Explorer Club

A community of people learning AI by building it, running it and testing it honestly: open models, self-hosted tools, and evaluations that report where things fail as well as where they work.

What we care about

• 🛠️ Hands-on: fine-tune, run and ship real models, not just read about them.
• 🖥️ Runs on our own machines: open weights and self-hosted tools wherever possible.
• 📏 Honest evaluation: fix the test before the run, keep a sealed holdout, compare against simple baselines, and publish the losses too.
• 🤝 Open by default: share code, models and lessons under permissive licences.

From our members

• IndiaVol by chelvaser: an open realized-volatility forecaster for NSE indices and F&O stocks. Chronos-2 fine-tune, pre-registered and tested on a sealed 2025–26 holdout, where it beats the classic HAR model on all five indices. Code · Collection

Join in

Members can publish models, datasets and Spaces under this organisation. Got a project you're proud of, or one that failed in an interesting way? Share it here.