Epsilon Health
Software Engineer - ML Infrastructure
San Francisco, CA · On-site · Full-time · Posted Oct 2, 2026
- Experience
- 6+ years in large-scale distributed systems or infrastructure; 2+ years in production ML infrastructure or systems.
About the job
Designs and operates distributed ML infrastructure for training, reinforcement learning, inference, and evaluation of large medical-imaging models, from research experimentation through production deployment.
What you'll do
- Build distributed foundation-model training infrastructure, including data loading, preprocessing, long-context parallelism, and checkpointing for volumetric imaging.
- Develop reinforcement-learning infrastructure for rollout generation, reward-model serving, and experience collection.
- Partner with researchers to productionize ideas and support model deployment, monitoring, and evaluation.
What they ask for
- 6+ years designing, building, and operating large-scale distributed systems or production infrastructure; 2+ years building production ML infrastructure or systems.
- Strong Python skills and expertise in PyTorch or JAX.
- Deep experience operating Kubernetes and cloud infrastructure at scale, and experience with distributed training at scale (FSDP, DeepSpeed, or Megatron-style parallelism).
- Experience supporting ML research workflows and prior technical leadership or mentoring experience.
Nice to have
- Experience with reinforcement-learning infrastructure, high-performance inference and serving, or training-performance optimization.
- Experience in a startup-like environment, building research experimentation platforms, or working with vision-language or multimodal architectures.
Tools
Python, PyTorch, JAX, Kubernetes, FSDP, DeepSpeed, Megatron, vLLM
AI focus
Enables large-scale medical-imaging foundation-model training and post-training, including reinforcement learning, inference, and evaluation for clinical deployment.
Model-written summary · the original posting is authoritative