Machine Learning Research Engineer

Our client, a Series B HealthTech startup, is looking for a Machine Learning Research Engineer to join their team.

As a Machine Learning Research Engineer, you'll design, build, and validate the models and pipelines at the core of our product. This role sits at the seam between research and production: one day you're framing a clinical prediction problem and choosing an evaluation that a physician would actually trust, the next you're building the data pipeline and monitoring that keeps that model safe and stable in the real world. You'll own problems end to end, work directly with clinical and product partners, and see your work show up in care that patients receive.

About the Role

• Frame clinical and operational problems as well-posed ML tasks, with labels and metrics that hold up to real-world and regulatory scrutiny.
• Build and evaluate models across the data types healthcare throws at you — structured EHR data, clinical text, waveforms or imaging, and increasingly LLMs and agentic systems for clinical workflows.
• Design rigorous evaluation and validation: subgroup and fairness analysis, calibration, robustness to distribution shift, and honest handling of confounding and label noise.
• Develop production-grade data pipelines and model infrastructure, with the monitoring and versioning that safety-critical deployment demands.
• Stress-test models for failure modes — bias across populations, data leakage, silent drift — and close the gaps you find before they reach patients.
• Collaborate closely with clinicians, product, and regulatory/quality partners to translate fuzzy clinical requirements into precise, measurable systems.

About You:

• Strong engineering fundamentals in Python and a deep-learning framework (PyTorch preferred), plus the ability to write clean, performant, production-quality code.
• Hands-on ML experience across the full loop — data pipelines, training, evaluation, and debugging systems that fail in subtle ways.
• A working understanding of modern ML applied to real, imperfect data: how to handle missingness, noisy labels, class imbalance, and distribution shift.
• Rigor and taste around evaluation: you care whether a metric measures what it claims to, and you notice when strong offline numbers won't survive contact with the real world.
• Sound judgment about safety, privacy, and the stakes of getting it wrong in a clinical setting.
• A bias toward shipping, and toward owning ambiguous problems without waiting to be told exactly what to build.

Nice to have

• Experience with healthcare or biomedical data (EHR/claims, clinical NLP, medical imaging, physiological signals) and standards like FHIR or OMOP.
• Familiarity with LLMs and agentic systems for clinical or documentation workflows.
• Exposure to regulated ML — HIPAA, model validation, or FDA/SaMD pathways.
• Publications or open-source work in ML, clinical ML, or evaluation.