Research Engineer, Clinical Logic (Healthcare AI)

An early-stage healthcare AI company turns written clinical guidelines into structured, checkable logic that runs at the point of care, and its system is already used inside a public health system. It is hiring a research engineer to explore how far frontier language models can go at reading, structuring and verifying medical reasoning, then to put the best results into production.

The work

• Build the pipeline that turns guideline documents into executable modules, covering document parsing, transformation and validation
• Shape and refine the intermediate representation used to encode guideline rules
• Run controlled experiments comparing open and proprietary models, fine-tuning and constrained generation
• Help define evaluation sets, metrics and error analysis that measure clinical faithfulness
• Create adversarial test cases, and write rule-based checkers using techniques such as program analysis, solvers or formal verification that surface vague wording, contradictions and missing guidance in the source material
• Meet regularly with the founding team and with the physicians who sign off on what the system produces

Who thrives here

• Three or more years of research or industry experience, or a PhD or postdoc in the final year
• A track record of shipping ML or NLP systems to production rather than only prototypes
• Production-grade Python and PyTorch, plus the Hugging Face Transformers ecosystem
• Hands-on experience evaluating, training or prompting large language models
• A degree in computer science, machine learning, math or a related field
• Clear writing about failed experiments and open uncertainty
• Experience at a frontier AI lab is expected; a background in compilers, formal methods or regulated domains is a plus

Fully remote for candidates based in Canada, with in-person team gatherings several times a year. Not able to sponsor or take on work visas.