AI / Machine Learning Engineer (LLM Applications)

A US healthtech client is moving its clinical documentation assistant from prototype to production and needs an engineer who has actually shipped LLM features rather than demoed them. You will own retrieval quality, evaluation, and the guardrails that keep a clinical product safe.

This is a build role, not a research role. The team cares about latency, cost per request, and whether the eval suite catches regressions before users do.

What you will do

• Design and ship RAG pipelines that hold up on real clinical documents|Build the evaluation harness that decides whether a prompt change is an improvement|Cut latency and cost per request without losing answer quality|Implement guardrails and fallback behaviour for a regulated product|Work directly with the client product and clinical teams on failure cases

Requirements

• Five or more years in software engineering, at least one shipping LLM features to production|Strong Python and solid API design instincts|Hands-on experience with retrieval, chunking strategies and vector databases|You have built an eval suite, not just eyeballed outputs|Fluent written and spoken English|Four hours or more of daily overlap with US Eastern time

Nice to have

• Exposure to HIPAA or other healthcare compliance requirements|Fine-tuning or distillation experience|Background in agent architectures and tool calling

Tech stack

Python, PyTorch, LangChain, pgvector, FastAPI, AWS Bedrock

← See all open roles