Senior ML Infrastructure Engineer (Medical Imaging AI)

A tiny, venture-backed healthcare startup is building an AI-assisted radiology service on top of multimodal models, and it needs an infrastructure engineer to make large-scale model training fast and dependable. The goal is for researchers to spend their time on science instead of fighting system limits.

Engineering scope

• Collaborate with researchers to learn how they work and plan for how their requirements will shift
• Build the multi-node training platform for large foundation models trained on sizable medical-imaging collections, including parallelism for long contexts plus checkpointing for volumetric CT and MR training
• Create fast data-loading and preprocessing pipelines that keep GPUs fed with big multimodal and volumetric data
• Prototype research ideas and turn them into production code, owning delivery from first experiment to release and ongoing monitoring
• Help with the pipelines that serve and deploy models in production, including staged rollouts, canaries and monitoring, working with the backend engineers
• Assemble the RL training stack by generating rollouts, serving reward models and gathering experience, so researchers can run large online RL with several reward signals

Required background

• Six to fifteen years as an ML infrastructure or distributed systems engineer
• Generalist range across serving, Kubernetes, data pipelines, training, RL and evaluation
• Large-scale distributed training with PyTorch or JAX, plus reproducible RL or online-learning loops with logging, checkpointing and evaluation
• Production model serving on Kubernetes and Docker, with GPU scheduling, autoscaling, versioning, canary rollouts and monitoring
• Time at a company with a strong ML infrastructure culture, ideally running real-time inference at scale
• Plus: staff-level or mentoring experience, and familiarity with experimentation platforms for ML models

On-site in the San Francisco Bay Area. Visa transfers and new visa sponsorship considered.