ML Engineer

The role

Build AI systems for healthcare workflows where output quality, evidence, latency, and cost all matter. The role spans retrieval over medical records, agent development, evaluation, and production engineering. Recent work has focused on improving appeal-letter quality while reducing the cost of the denials system.

‍

What you’ll do

• Improve agent quality. Investigate whether failures come from missing knowledge, failed retrieval, or model reasoning; test changes against real cases and review complete outputs, not just intermediate scores.
• Build retrieval and context systems. Work on programmatic medical-record file systems, specialized retrievers, and agent state so models can find and use the right evidence.
• Make quality and cost measurable. Add tracing and accurate model-level cost attribution; build reusable benchmarks from production cases before optimizing or changing models.
• Experiment with new open weights models and training. Evaluate model-and-harness combinations, synthetic data, fine-tuning, and reward methods where they can improve cost-performance without sacrificing quality.
• Ship reliable systems. Turn experiments into maintainable services, integration tests, observability, and documented work that teammates can build on.

‍

What we’re looking for

• Strong Python and software-engineering fundamentals; you can explain, debug, and improve code you wrote, including code developed with AI assistance.
• First-principles reasoning about LLM behavior, algorithms, retrieval choices, and production tradeoffs. Depth of reasoning matters more than prior ML credentials alone.
• A habit of running controlled evaluations, inspecting failures against source evidence, and protecting quality when changing models or architecture.
• Collaborative problem-solving, clear technical communication, and an interest in following relevant AI research and applying it pragmatically.