Machine Learning Engineer

Latent•San Francisco, CA

2d•$225,000 - $300,000•Onsite

About The Position

About Latent Health Healthcare today is only truly personalized for two groups: those with wealth and access, and those with physicians in their immediate family. For everyone else, care is fragmented and impersonal. Medical history is scattered across systems that don’t communicate. Physicians have minutes to understand decades of context. And when something goes wrong, patients are left with tools that understand medicine broadly—but not the individual. We believe this can be fundamentally rebuilt. At Latent Health, we are building systems that understand both: the population (clinical knowledge at scale) and the individual (longitudinal patient history) Our models are designed to answer complex clinical questions with patient-specific context and verifiable reasoning. Our dataset represents one of the most clinically diverse populations in the United States, including patients with chronic illness and complex disease. Each patient record contains extraordinary depth. ML at Latent Health The Machine Learning team is responsible for building systems that run in real clinical workflows. We work on: Verifiable reinforcement learning at scale Mid-training and post-training of foundation models Novel objectives derived from longitudinal patient data We are a small group of researchers and engineers focused on pushing the frontier while shipping real systems into production. We are a small team and expect engineers to take ownership of critical systems, not components. The Role As a Machine Learning Engineer, you will own the design, development, and operation of production-grade ML systems that run in real clinical workflows. You will drive systems from ambiguous problem definition through to reliable production deployment, setting technical direction along the way. We are primarily hiring for senior and staff-level engineers who are comfortable owning critical systems end-to-end. This role involves owning systems that directly impact real patient outcomes.

Requirements

Strong foundation in machine learning and software engineering
Track record of building and owning ML systems in production where performance, reliability, or correctness materially mattered
Experience driving ambiguous ML problems from 0→1, including problem formulation, model design, and productionization
Hands-on experience with PyTorch or similar frameworks
Ability to operate independently in high-ambiguity environments with minimal guidance
Strong product and engineering judgment — you know when to use ML, when not to, and how to scope problems accordingly
Comfort working in a fast-moving, early-stage environment
Experience working on systems where decisions have real-world consequences (e.g., healthcare, finance, infrastructure)

Nice To Haves

Experience deploying LLMs in production environments
Experience building distributed systems or large-scale data pipelines
Experience working with clinical, biomedical, or other regulated datasets

Responsibilities

Own end-to-end ML systems, including architecture, data, modeling, evaluation, and production infrastructure
Train and fine-tune large language models (LLMs) for:
Clinical reasoning
Medical question answering
Evidence-grounded generation
Make and own tradeoffs across accuracy, latency, cost, and safety in high-stakes production environments
Develop evaluation frameworks to ensure model safety and clinical validity
Integrate ML systems into product workflows and patient-facing applications
Monitor system performance in production and iterate based on real-world usage and feedback
Define what “correct” means in ambiguous clinical workflows in collaboration with engineers and clinicians