Machine Learning Infrastructure Engineer

San Francisco, CA On-site (5 days/week) Full-time Compensation: $200K–$400K + competitive early-stage equity About The Company Our client is a Series A AI research lab building large-scale foundation models for scientific and physical-AI domains. Backed by top-tier investors, they are pursuing a deliberately non-consensus technical thesis and are among the best-funded teams in their space. The founding team comes from self-driving, robotics, and scientific research, and they are scaling their research and engineering org significantly this year. Founded 2024 Small, fast-growing team Industry: AI / foundation models / physical AI The Role You would own the distributed training and inference backbone for a foundation model trained from scratch — standing up clusters, building data and training pipelines at petabyte scale, and squeezing performance out of GPUs at a low level across model scales. What You'll Be Doing Design, deploy, and maintain large distributed ML training and inference clustersBuild efficient, scalable end-to-end pipelines to manage petabyte-scale datasets and training across the full ML lifecycleResearch and test training approaches, including parallelization techniques and numerical-precision trade-offs across model scalesProfile and debug low-level GPU operations to optimize performanceTrack new research and bring fresh ideas into the work Tech stack: Distributed training frameworks (FSDP, DeepSpeed), NVIDIA GPUs, Linux, Python, C++, Kubernetes/Docker, and a major cloud platform (GCP, AWS, or Azure). Requirements 2–10 years building large-scale ML infrastructure for core foundation modelsHands-on experience building infrastructure for foundation models trained from scratch, rather than fine-tuning existing modelsA background at a science-focused or physical-AI company (for example self-driving, robotics, or biology)Deep, demonstrable expertise optimizing large-scale training and inference workloadsWorking proficiency with distributed training frameworks such as FSDP or DeepSpeedA clear pattern of intentional, mission-driven career decisionsAble to work on-site 5 days/week in San Francisco (relocation supported) Nice to Haves Generalist experience spanning the full ML lifecycleLow-level GPU performance optimization and debugging (CUDA, JAX) Why Join Take a bet on a distinctive, non-consensus approach to building intelligenceJoin early, with real ownership of the training and inference backboneWork in a domain with fast, objective ground-truth feedback and data at a scale beyond typical LLM trainingWell-funded and building a strong, senior research and engineering team Details Location: San Francisco, CAWork policy: In-person 5 days/week (relocation supported)Compensation: $200K–$400K + competitive early-stage equityVisa sponsorship: Open to supporting work authorization for the right candidateEmployment type: Full-time