Compiled Health - Member of Technical Staff, Compiler (Remote)
- Remote, Canada
Remote
Applications open in a new tab.
Job details
- Workplace
- Remote
- Experience
- 3+ years of research or industry experience, or final-year PhD / postdoc
- Posting language
- English
- Posted
Original job description
Source: Paraform - https://www.paraform.com/share/compiled-health/cmtsxkxh100150cl7imvz606z
Compensation: $250,000-$400,000 base + competitive equity (up to 5% for a frontier-lab candidate, from a 15% pool; cash/equity ratio flexible) + health, dental, vision, flexible PTO, paid parental leave.
Location: Remote - US or Canada, with offsites several times a year. Visa: not open to visas (possible exceptions for frontier-lab employees).
About this role
Research role with production stakes: 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 clinical compiler, which is live in Canadian public health care today. Ideal for someone who has published in clinical AI, wants out of the demo cycle, and believes clinical decision logic must be inspectable.
What you'll do
Build the pipeline converting clinical standards into executable modules: parsing with document and layout models, transformation, validation
Design and iterate the target syntax/representation for guideline logic (exclusions, precedence, recommendations)
Run experiments: open vs closed models, fine-tuning, constrained decoding
Co-design benchmarks and adversarial cases measuring source fidelity, logical completeness and recommendation correctness; error analysis
Develop deterministic validators using static analysis, constraint solving or formal methods to surface ambiguity, contradiction and source silence
Work directly with the founders and the clinicians who validate the outputs; turn model failures into hypotheses and better system design
Tech stack
Python, PyTorch, Transformers, LLMs, Static Analysis, SMT/SAT Solvers, FHIR, HL7 v2
About Compiled Health
Clinical AI research and formalization company compiling the standard of care into deterministic, evidence-traceable decision artifacts; compiler live in Canadian public health care. Founded by a physician who worked with OpenAI's health safety research team. ~6 people, founded 2024, $3M seed plus a committed billionaire family office.
3+ years of research or industry experience, or final-year PhD / postdoc
Published in clinical AI (strongly preferred)
Python, PyTorch, Transformers/LLMs; static analysis, SMT/SAT solvers; FHIR / HL7 v2 a plus
US or Canada based; US citizen/Green Card (visa exceptions possible for frontier-lab employees)
More about this job
Responsibilities
You will build pipelines to convert clinical standards into executable modules and design the target syntax for guideline logic. Additionally, you will run experiments with various models and develop deterministic validators to ensure recommendation correctness.
Requirements
Candidates should have at least 3 years of research or industry experience, or be a final-year PhD/postdoc. Proficiency in Python, PyTorch, and LLMs is required, with a strong preference for those who have published in clinical AI.
Skills
- Python
- PyTorch
- Transformers
- LLMs
- Static analysis
- SMT solvers
- SAT solvers
- FHIR
- HL7 v2
- Clinical AI
- Constraint solving
- Formal methods
- Data parsing
- System design
- Clinical logic
Benefits
- Competitive equity
- Health insurance
- Dental insurance
- Vision insurance
- Flexible PTO
- Paid parental leave
Remote locations
- United States
- Canada
Remote locations (standardized)
- Canada
- United States
Visa sponsorship
Not detected in the job text
Categories
- Technology
- Software
- Science & Research
- Healthcare
- Data & Analytics
Keywords
- Clinical AI
- Compiler
- Python
- PyTorch
- Transformers
- LLMs
- Static Analysis
- SMT Solvers
- SAT Solvers
- FHIR
- HL7 v2
- Clinical Logic
- Formal Methods
- Constraint Solving
- Healthcare
- Research
- Evidence-traceable
- Data Parsing
- System Design
- Adversarial Testing
- Fine-tuning
- Constrained Decoding
- Clinical Standards
- Executable Modules
- Validation
- Source Fidelity
- Logical Completeness
- Recommendation Correctness
- Error Analysis
- Deterministic