Healthcare AI/ML Engineer (LLM, RAG, Health Data) - Individual Only

We are working with a precision health company on its AI architecture and roadmap. We are bringing on one engineer to do the hands-on technical work on this engagement alongside us. This starts small, about 5 to 10 hours a week, and should grow as the work moves from review into build. I'm looking for someone who wants to stay with it long term. WHAT YOU'LL DO - Review an LLM-based health assistant: how it retrieves scientific evidence, how it reaches a user's health data, and how its answers are checked - Review and improve the data model for health data collected over time from several sources (lab results, genomic data, wearables) - Build test sets and automated checks for LLM answers, including guardrails around health claims - Work through privacy and access questions: PHI handling, role-based access, audit trails, de-identification - Write up findings as short, clear assessments and recommendations - Later on, build some of what we recommend (Python, SQL, AWS) You'll own this work day to day. MUST HAVE - Healthcare or life-science experience on a real project: clinical, lab, genomic, wearable or EHR data. This is required. Please don't apply without it. - LLM applications in production: RAG, tool/function calling, prompt design, and above all how you tested them - Strong data modeling in SQL/Postgres, and solid Python - Working knowledge of HIPAA and of handling sensitive health data - Clear written English. A lot of this job is writing things a founder and a small technical team can act on. - You are an individual freelancer and will do all of the work yourself. No agencies, no teams, no passing work to someone else. NICE TO HAVE - Bioinformatics background (genotyping arrays, DNA methylation, biological age models) - LOINC, FHIR or other health data standards - Statistics for repeated measurements over time and for observational data - Able to read Go HOW THIS WORKS - Hourly through Upwork, $15 to $25 depending on experience - At least 5 hours of overlap with US Pacific time for a daily check-in. Everything else is async. TO APPLY Start your proposal with the word "biomarker" so I know you read this. Then answer the questions below in your own words. Short, specific answers beat long ones.