All work
Clinical AI
Northfield Health
2026
A retrieval-grounded assistant that drafts triage notes for emergency nurses, checked against 40,000 past cases before anything reaches a chart.

Scroll to resolve
−41%
Time to first triage note
Client
Northfield Health
Role
Lead ML engineer
Stack
RAG · Llama 3 · pgvector · FastAPI
Year
2026
The problem
Emergency nurses were spending the first eleven minutes of every intake typing notes that mostly repeated what the patient had just said.
What I built
I built a retrieval pipeline over 40,000 anonymised past cases, so every suggested sentence cites a similar, real precedent. A small classifier refuses to draft when confidence is low, and nurses edit rather than approve.
The result
After a three-month pilot across two hospitals, time to first note dropped by 41%, with zero drafts flagged by the clinical safety review.