Job description

ABOUT US Epsilon Health is transforming medical imaging and diagnostics by bringing frontier AI into clinical practice. We're tackling some of healthcare's most critical challenges, building technology that directly improves patient outcomes. This year, we've served over 300,000 patients and now serve 3,000+ daily, working with some of the largest medical institutions in the U.S. By year-end, we're on track to interpret 1% of all daily U.S. X-rays. Our research team brings experience from DeepMind and Meta, powered by one of the industry's most comprehensive proprietary medical imaging datasets, continuously enriched by real-world clinical data. Together, we're building a future where frontier AI helps save millions of lives today, not tomorrow. Recently out of stealth, we're proudly backed by Jack Altman, Uncork Capital, Renegade Partners, and AlleyCorp, one of NYC's most active healthcare venture firms. ROLE OVERVIEW We’re looking for an ML infrastructure engineer to design and build the core systems that enable scalable, efficient training of large models for deployment and research. Your goal is to make experimentation and training at Epsilon Health fast and reliable to ensure our research teams can focus on science rather than system bottlenecks. Sitting in the Engineering team and working closely with research, you'll own the distributed training and reinforcement learning infrastructure our foundation-model and post-training work runs on, and the inference and evaluation systems that carry models from experimentation into production. KEY RESPONSIBILITIES - Partner directly with researchers to deeply understand their workflows, then anticipate and design for how those needs will change - Build a distributed training infrastructure for foundation models on large-scale medical imaging, including the long-context parallelism and checkpointing that volumetric CT/MR training demands. - Build high-throughput data loading and preprocessing that keeps GPUs saturated on large volumetric and multimodal datasets. - Partner with researchers to prototype new ideas and translate them into production-ready code, owning end-to-end delivery from experimentation through deployment and monitoring. - Contribute to production serving and deployment pipelines (model rollout, canary deployments, and monitoring) alongside the backend team. - Build the reinforcement learning training stack (high-throughput rollout generation, reward-model serving, and experience collection), enabling the research team to run online, multi-reward RL at scale. QUALIFICATIONS - 6+ years of experience designing, building, and operating large-scale distributed systems or infrastructure in production - Have 2+ years of experience building ML infrastructure or systems in production - Strong Python skills and expertise in PyTorch or JAX - Experience and familiarity with the compute, tooling, and workflow needs of large-scale machine learning research - Experience building infrastructure or platforms specifically for research or machine learning workflows - Deep experience building and operating Kubernetes and cloud infrastructure at scale - Experience with distributed training at scale (FSDP, DeepSpeed, or Megatron-style parallelism) and the systems concerns of keeping large GPU jobs efficient - Prior experience as a technical lead or mentor for other engineers PREFERRED QUALIFICATIONS - Experience operating in a startup or startup-like environment, i.e. a small, fast-moving team with high autonomy - Experience building reinforcement learning training infrastructure: rollout generation, reward-model serving, or online/off-policy learning systems - Experience with high-performance inference and serving (vLLM, SGLang, TensorRT, or Triton) for both training-time rollouts and production - Experience optimizing inference and serving for large models: batching, KV/prompt caching, quantization, and low-latency, high-throughput sampling. - Experience optimizing training performance: parallelism, distributed communication, mixed/low precision, and utilization. - Experience building internal training or experimentation platforms used by research teams, supporting A/B testing and experimentation workflows - Familiarity with vision-language models (VLMs) or multimodal architectures

Requirements

Must-haves

Partner directly with researchers to deeply understand their workflows, then anticipate and design for how those needs will change
Build a distributed training infrastructure for foundation models on large-scale medical imaging, including the long-context parallelism and checkpointing that volumetric CT/MR training demands.
Build high-throughput data loading and preprocessing that keeps GPUs saturated on large volumetric and multimodal datasets.
Partner with researchers to prototype new ideas and translate them into production-ready code, owning end-to-end delivery from experimentation through deployment and monitoring.
Contribute to production serving and deployment pipelines (model rollout, canary deployments, and monitoring) alongside the backend team.
Build the reinforcement learning training stack (high-throughput rollout generation, reward-model serving, and experience collection), enabling the research team to run online, multi-reward RL at scale.
6+ years of experience designing, building, and operating large-scale distributed systems or infrastructure in production
Have 2+ years of experience building ML infrastructure or systems in production
Strong Python skills and expertise in PyTorch or JAX
Experience and familiarity with the compute, tooling, and workflow needs of large-scale machine learning research
Experience building infrastructure or platforms specifically for research or machine learning workflows
Deep experience building and operating Kubernetes and cloud infrastructure at scale
Experience with distributed training at scale (FSDP, DeepSpeed, or Megatron-style parallelism) and the systems concerns of keeping large GPU jobs efficient
Prior experience as a technical lead or mentor for other engineers

Nice-to-haves

Experience operating in a startup or startup-like environment, i.e. a small, fast-moving team with high autonomy
Experience building reinforcement learning training infrastructure: rollout generation, reward-model serving, or online/off-policy learning systems
Experience with high-performance inference and serving (vLLM, SGLang, TensorRT, or Triton) for both training-time rollouts and production
Experience optimizing inference and serving for large models: batching, KV/prompt caching, quantization, and low-latency, high-throughput sampling.
Experience optimizing training performance: parallelism, distributed communication, mixed/low precision, and utilization.
Experience building internal training or experimentation platforms used by research teams, supporting A/B testing and experimentation workflows
Familiarity with vision-language models (VLMs) or multimodal architectures

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aimlpythonmachine learningkubernetes

Platform: ashby
Effort: quick

Benefits

Benefits detected in the listing.

Vision