Member of Technical Staff, Compiler (Remote)
Base salary range of $250K–$400K depending on experience, plus competitive equity and benefits including health, dental, vision, flexible PTO, and paid parental leave. Equity goes up to 5% for a candidate coming out of a frontier lab, granted from a 15% option pool. The ratio between cash and equity is flexible depending on what the candidate wants.
Compiled Health compiles the standard of care into decision artifacts that execute in front of a patient: schema-enforced, verified for exhaustiveness, and traceable to the source page that justifies them. The compiler is the heart of it, and it is live in Canadian public health care today. This is a research role with production stakes. You will study how far frontier models can be pushed in extracting, structuring, and proving clinical logic, design the experiments that settle what works, and ship the winning approach into the pipeline.
You will feel at home here if you have published in clinical AI, want out of the demo cycle, and think the logic behind a clinical decision has to be inspectable.
What You'll Do
Build the pipeline. Convert clinical standards into executable modules: parsing with document and layout models, transformation, validation.
Design the target syntax. Iterate on the representation for guideline logic.
Run the experiments. Open versus closed models, fine-tuning, constrained decoding.
Co-design the benchmarks. Datasets, metrics, and error analysis for clinical fidelity.
Work with the people who validate it. Directly with the founders and the clinicians who stake their license on the outputs.exclusions, precedence, and recommendations.
Build benchmarks and adversarial cases that measure source fidelity, logical completeness, and recommendation correctness.
Turn model failures into hypotheses, experiments, and better system design — improving the compiler that runs in clinical production.
Develop deterministic validators using static analysis, constraint solving, or formal methods to make ambiguity, contradiction, and source silence explicit.
Compiled Health is a clinical AI research and formalization company building a future where the standard of care is executable — where what medicine knows is what medicine does, at every visit, in every setting.
The accumulated, evidence-graded knowledge of medicine still lives in documents built to be read, not executed. Compiled Health compiles that knowledge into decision artifacts that are measurable against the standard, deterministic where they must be, and traceable to their source evidence. The compiler is live in clinical production today: clinicians in public health networks act on its recommendations every day.
Founded by a physician who previously worked with OpenAI's health and safety research team, Compiled Health operates on a core thesis that separates AI usage at build time from runtime — meaning clinicians interact with deterministic, validated outputs rather than raw generative AI. This approach addresses the fundamental challenges of applying generative AI to tightly regulated, high-stakes clinical environments.
The company operates a dual-entity structure: Compiled Health serves as the intelligence and research arm with a frontier lab flavor, while a companion distribution entity delivers the research pipeline as a traditional SaaS product to healthcare organizations.
Backed by a $3M seed round from VCs in Canada and a billionaire family office committed to supporting the company through Series A and beyond, Compiled Health is positioned for long-term research-driven growth with a substantial runway. The team is small, deeply technical, and fully remote across North America, meeting in person multiple times per year for offsites.
Team size
Founded
Website
compiledhealth.com/Total funding
Company locations
Compiled Health is a team of six people today — small, focused, and deeply mission-driven. The company was co-founded by Antoine Denis, a physician by background who practiced family medicine before spending two years working closely with OpenAI's health safety research team, and Christian, the CEO who leads all research efforts.
Working at Compiled Health means:
Frontier-caliber research with real-world impact: Your work directly affects clinical care in public health networks today. This isn't theoretical — the compiler is live in production.
Startup intensity: The team values high-output, high-ownership contributors who are excited to push hard and move fast. This is not a 35-hours-a-week environment.
Fully remote, North America: The team is distributed across the US and Canada (co-founders are based in Montreal and Vancouver), with in-person offsites every couple of months to build connection and collaborate deeply.
Direct access to founders: With a team this small, every hire works directly with the co-founders and has outsized influence on the company's direction and culture.
Long runway: Backed by committed investors including a billionaire family office, the team can focus on doing exceptional research and building something meaningful without the typical early-stage cash crunch anxiety.
Compiled Health's core product is a clinical compiler — a system that transforms clinical standards of care (which arrive as prose, tables, algorithms, footnotes, exceptions, and cross-references) into executable, evidence-bound decision artifacts.
The key innovation is the separation of AI at build time versus runtime:
- At build time: AI models do their heaviest work — proposing clinical logic, measuring where they fail, and converting successful outputs into versioned artifacts. This is where research and engineering intersect.
- At runtime: When clinicians actually use the platform, there is no generative AI. Instead, clinicians interact with deterministic, validated decision artifacts that are measurable against the standard and fully traceable to source evidence.
This architecture directly addresses the risks of applying generative AI to clinical care — lawsuits, hallucinations, and the fundamental mismatch between probabilistic AI outputs and the precision required in a tightly regulated, safety-critical domain.
The compiler is live in clinical production today, with clinicians in public health networks acting on its recommendations daily. The system includes deterministic validators using static analysis, constraint solving, and formal methods to make ambiguity, contradiction, and source silence explicit.