Senior Applied Scientist

Hippocratic AI • Seattle, WA

16h

About The Position

Hippocratic AI is building a recursive self-improvement system: a machine learning system that iteratively improves itself through feedback, evaluation, and automated learning loops. The Senior Applied Scientist will help build the engineering pipeline that keeps these loops fast, reliable, and trustworthy, including training and evaluation pipelines, reward and feedback signals, and safeguards against silent degradation or objective gaming. This is an engineering-first role requiring deep reinforcement learning knowledge, production ML engineering skills, and the ability to reason about reward design, credit assignment, and instability in feedback-driven systems.

The role focuses on building reproducible, reliable training, evaluation, and deployment systems; designing feedback and reward signals; developing evaluation harnesses and metrics; maintaining data pipelines and automated data flywheels; diagnosing model-quality regressions; and stabilizing non-stationary training and feedback loops. The scientist will collaborate with research and product teams to turn methods into robust, shippable systems.

Hippocratic AI describes itself as building a healthcare-only, safety-focused LLM intended to transform patient outcomes globally. The company is co-founded by CEO Munjal Shah and a team of healthcare and technology leaders, and is backed by healthcare and AI investors.

Requirements

• Excellent Python skills and ability to write clean, well-tested ML training code.
• Solid understanding of data pipelines, distributed or large-scale training, and experiment tracking.
• Ability to diagnose why a model's quality has silently declined, not just why it has crashed.
• Hands-on experience with at least one feedback or learning loop, such as building or owning a reward model, evaluation harness, or data pipeline for an RLHF/RLAIF or active-learning system.
• Experience running a retraining or continual-learning pipeline where a model consumed its own predictions or production data, such as in ranking, recommendations, fraud, or spam.
• Experience fine-tuning LLMs with human or AI feedback, or building agentic evaluation harnesses.
• Working knowledge of reward modeling, on-policy versus off-policy tradeoffs, and credit assignment.
• Ability to identify or reason clearly about feedback-system failure modes, including reward hacking, specification gaming, and feedback loops that amplify errors.
• Comfort evaluating non-stationary systems whose behavior and data distribution change over time.
• A measurement-first approach that treats evaluation as a first-class deliverable.
• Experience shipping an ML system to production and maintaining it over time.

Nice To Haves

• A PhD or MS in reinforcement learning or machine learning paired with real production experience.
• Experience at a lab or company working on RLHF, agents, or large-scale ML infrastructure.
• Familiarity with LLM fine-tuning, evaluation frameworks, or agent orchestration.
• Experience building or shipping an end-to-end system that improved from its own outputs or feedback, such as RLHF/RLAIF pipelines, self-play systems, active-learning loops, automated data flywheels, or agentic evaluation harnesses.

Responsibilities

• Build and maintain the training, evaluation, and deployment loops at the core of the self-improvement system, emphasizing reproducibility and reliability.
• Design and implement reward and feedback signals, and investigate and mitigate reward hacking, specification gaming, and distribution drift.
• Build evaluation harnesses and metrics before developing models.
• Own data pipelines and automated data flywheels that feed the learning loop.
• Debug subtle model-quality regressions and stabilize training and feedback loops that become non-stationary.
• Collaborate with research and product teams to turn methods into robust, shippable systems.

What This Job Offers

• Job Type: Full-time
• Career Level: Senior
• Education Level: No Education Listed