Bravim Purohit
M.S. Computer Science · Illinois Institute of Technology · May 2026
I build AI systems that hold up in the real world — ambient scribes, agentic pipelines, fine-tuned clinical LLMs, and whatever infrastructure it takes to make them run. Three years of shipping in healthcare, where bugs have real consequences and good evaluation is non-negotiable. Currently at a stealth medtech startup in SF. Open to founding-engineer and ML roles at applied AI startups where the work still matters.
“The best way to predict the future is to invent it.”
— Alan Kay
Compact summary. Detailed engagement notes on request, in interviews.
Three things on my plate right now. One is under NDA, two are public.
HIPAA-grade systems for healthcare workflows — the kind where a bug isn't just annoying, it affects someone's care. Can't say more publicly, but happy to talk in an interview.
Can frontier AI agents actually handle the administrative work that burns out clinicians? HealthAdminBench (ICLR 2026, Stanford collab) is our attempt to measure that rigorously.
Every design decision is documented and posted. Claude Code as the IDE. 30 days to go from scaffold to something I am proud of.
Public projects. Company work is reserved for interviews, kept off the public site by design.
Real-time ambient scribe for clinical conversations. WebSocket speech-to-text for live transcription, incremental SOAP note generation, and a Risk Sentinel module that surfaces clinical risks with transcript-backed evidence. Non-blocking LLM architecture keeps transcription alive even if generation fails.
Clinical retrieval-augmented generation pipeline combining text and imaging embeddings for case-aware diagnostic insights. AstraDB for vector retrieval, MedLLaMA for grounded report generation.
Built end-to-end with Claude Code skills. Each design decision is documented and posted as a daily thread.
Two more applied AI projects shipping during the 30-day build-in-public arc.
Pulled live from the GitHub API at build time. Last 52 weeks of activity.
A supporting credential for the applied work above. Full author lists and citations on arXiv and in the journal.
Co-author. Stanford University collaboration.
A benchmark evaluating frontier Computer-Use Agents (Claude, GPT) on real healthcare administration workflows: prior authorization, appeals, and DME order processing. The paper shows where current CUAs break on production healthcare admin and what closing the gap requires.
Peer-reviewed journal article.
Peer-reviewed paper in Springer SN Computer Science. Awarded Best Engineering Innovation for the month of publication.