I build systems where AI agents do real work — and a human keeps the only irreversible button.
Currently building (private beta): a local-first operating system for AI agents — a fleet of specialized agents that review code, run QA, post verdicts, apply labels, and merge their own pull requests through human-gated pipelines, all running on hardware you own. Multi-model by design: different models write, review, and judge each other, because no single model's blind spots should go unchecked.
Recent things the fleet did without me: reviewed and shipped its own UI redesign end to end, audited every screen of its dashboard with a vision model, and started publishing a daily brief of its own work — deliverables, failures, and spend, every line traceable to a real run.
🌐 sajidpalagiri.com · Hyderabad, India