Senior Research Engineer

Job Description

Research Engineer, Clinical Reasoning

Location: New York City or San Francisco

Work Style: Hybrid Onsite

Employment Type: Full-time

Focus: Clinical AI, Post-Training, Reinforcement Learning, Evals, Retrieval, Agentic Reasoning

About Our Client

Our client is building a frontier AI lab focused on clinical reasoning.

The company operates a real clinical practice where patients interact with an AI doctor every day. This gives the team access to a unique feedback loop: real patient interactions, real clinical workflows, and real-world healthcare data that can be used to test, evaluate, and improve clinical AI systems over time.

The mission is to make high-quality healthcare more accessible through AI while continuously improving clinical safety, accuracy, reasoning quality, and trust.

About the Role

Our client is hiring a Research Engineer, Clinical Reasoning to help build the reasoning systems, learning methods, retrieval algorithms, post-training data pipelines, and evaluation infrastructure behind its clinical AI platform.

This role blends research and engineering. The right person can run experiments, post-train models, build high-quality data pipelines, debug ML systems, and ship production-quality infrastructure that helps clinical AI improve with every iteration.

This is not a pure research role and it is not a generic software engineering role. It is designed for someone with a strong spike in ML or AI research and the engineering ability to turn open-ended research ideas into working systems.

What You’ll Build

Agentic Clinical Reasoning

\n
• Design and implement next-generation clinical AI reasoning systems
\n
• Build architectures for reasoning, reflection, verification, tool use, routing, uncertainty handling, and escalation
\n
• Create systems where specialized agents and models work together to support safe, reliable clinical decisions
\n
• Help clinical AI systems earn greater autonomy over time through better reasoning, evaluation, and feedback loops
\n

Evaluation and Measurement

\n
• Build evaluation platforms, rubrics, simulations, and experiments that measure clinical AI performance
\n
• Identify when a benchmark score rewards the wrong behavior
\n
• Determine whether improvements should come from reasoning, retrieval, model behavior, data, or engineering changes
\n
• Measure whether each intervention genuinely improves safety, accuracy, usefulness, and clinical reliability
\n

Models, Learning, and Post-Training

\n
• Run post-training experiments to improve clinical AI behavior
\n
• Apply methods such as fine-tuning, distillation, reinforcement learning, preference optimization, and prompt or system optimization
\n
• Build training data, feedback, reward, and experimentation pipelines
\n
• Partner with in-house researchers to translate training objectives into concrete data and evaluation specifications
\n
• Scale training infrastructure and experimentation workflows
\n
• Create and manage both real-world and synthetic data pipelines
\n

Search, Retrieval, and Grounding

\n
• Build search, ranking, retrieval, and grounding algorithms that connect clinical reasoning to trusted medical evidence, patient context, and partner-specific content
\n
• Improve retrieval based on downstream clinical decision quality, not just document relevance
\n
• Build systems that help clinical AI produce grounded, evidence-backed, trustworthy answers
\n

What We’re Looking For

\n
• Strong spike in ML or AI research, especially in areas like post-training, reinforcement learning, evaluations, interpretability, model behavior, or agentic systems
\n
• Strong software engineering ability with the ability to build reliable tools, debug systems, and ship working infrastructure
\n
• Experience post-training models and building or curating high-quality post-training data
\n
• Ability to design, evaluate, and improve data pipelines for model training and clinical reasoning
\n
• Strong judgment around data quality, task realism, evaluation design, and model behavior
\n
• Fast problem-solving and code comprehension skills, especially in unfamiliar codebases or ML pipelines
\n
• Comfort debugging PyTorch workflows, model training issues, data pipelines, and evaluation systems
\n
• Ability to translate research goals into concrete datasets, experiments, and measurement frameworks
\n
• Ability to work across research and engineering, turning open-ended AI problems into usable systems
\n
• Clear communication with research, engineering, clinical, product, and data teams
\n

Ideal Background

Successful candidates may come from backgrounds such as:

\n
• Applied ML research
\n
• Research engineering
\n
• Post-training or reinforcement learning
\n
• Model evaluation and interpretability
\n
• Data science with strong engineering depth
\n
• Software engineering with a research-oriented AI focus
\n
• Clinical AI, healthcare AI, or safety-critical AI systems
\n

The team cares more about depth of work, research judgment, engineering ability, and ability to ship than a specific credential or career path.

Bonus Experience

\n
• Published research, patents, meaningful open-source contributions, or novel production ML systems
\n
• Experience building AI systems at an early-stage or high-growth company
\n
• Experience in healthcare, clinical AI, or another regulated or safety-critical domain
\n
• Familiarity with clinical workflows, healthcare data, HIPAA, FHIR, EHR systems, or HL7
\n
• Experience with human-feedback systems, RLHF, simulation, or synthetic data generation
\n
• Experience in AI safety, bias detection, calibration, fairness, or model reliability
\n
• Experience building evaluation harnesses for LLMs, agents, or clinical decision systems
\n

Why This Opportunity

\n
• Join a frontier AI lab focused on clinical reasoning
\n
• Work with real patient interactions and real clinical workflows
\n
• Build systems that help clinical AI reason, learn, retrieve evidence, and improve over time
\n
• Own problems end-to-end across research, data, training, evaluation, and deployment
\n
• Partner with researchers, engineers, clinicians, and product teams on high-impact healthcare AI systems
\n
• Build in a safety-critical domain where evaluation, grounding, and reliability truly matter
\n
• Help define how clinical AI earns trust and autonomy over time
\n

Ideal Candidate Profile

The ideal candidate is a research-minded engineer with a real spike in ML, AI research, post-training, reinforcement learning, evaluations, or model behavior.

They can understand research objectives, build the data and evaluation systems behind them, debug ML pipelines, post-train models, and ship production-quality infrastructure. They have strong data taste, strong engineering ability, and the judgment to build clinical AI systems that are measurable, grounded, and safe.

This person wants to work at the intersection of frontier AI research and real-world healthcare.