Staff Compiler Engineer, Clinical AI Research

OpenKrill; United States; À distance; USD 250 000 - 400 000; Plein temps

Avantages offerts par ce poste

Health insurance

Dental insurance

Vision insurance

Flexible PTO

Paid parental leave

Résumé du poste

Paraform is seeking a research-focused professional to study how far frontier models can be pushed in extracting, structuring and proving clinical logic. You will design experiments that settle what works and ship the winning approach into the clinical compiler, already live in Canadian public health care today.

The role involves building deterministic validators, co-designing benchmarks, and working directly with the founders and clinicians to translate model failures into improved system

Qualifications

• 3+ years of research or industry experience, or final-year PhD/postdoc.
• Published in clinical AI (preferred).
• Proficiency with Python, PyTorch, Transformers/LLMs.
• Familiarity with static analysis, SMT/SAT solvers, FHIR/HL7 v2 a plus.
• US or Canada based; visa restrictions may apply for frontier-lab roles.

Responsabilités

• Design experiments to push frontier models for clinical logic.
• Build pipeline to convert clinical standards into executable modules.
• Collaborate with founders and clinicians to validate outputs.
• Develop deterministic validators to surface ambiguity and contradictions.
• Run experiments comparing open vs closed models, fine-tuning, and constrained decoding.

Connaissances

Research experience

Clinical AI familiarity

Python

PyTorch

Transformers/LLMs

Static analysis

SMT/SAT solvers

Formation

Final-year PhD / postdoc

3+ years of research/industry experience

Outils

FHIR

HL7 v2

Description du poste

Paraform is seeking a research-focused professional to study how far frontier models can be pushed in extracting, structuring and proving clinical logic. You will design experiments that settle what works and ship the winning approach into the clinical compiler, already live in Canadian public health care today.

The role involves building deterministic validators, co-designing benchmarks, and working directly with the founders and clinicians to translate model failures into improved system