SF ML Infrastructure Architect – Foundation Models

davidjoseph-co; San Francisco (CA); On-site; USD 200,000 - 400,000; Full time

Benefits offered by this job

Relocation support

Visa sponsorship

On-site in SF

Equity compensation

Job summary

Unknown Company in San Francisco, CA seeks a lead ML infrastructure engineer to own the distributed training and inference backbone for a foundation model trained from scratch. You will stand up clusters, build data pipelines at petabyte scale, and optimize GPU performance across model scales.

You will work with FSDP/DeepSpeed, NVIDIA GPUs, Linux, Python and C++, in a distributed cloud environment across GCP/AWS/Azure, with relocation supported.

Qualifications

• 2–10 years building large-scale ML infrastructure for core foundation models.
• Hands-on experience building infrastructure for foundation models trained from scratch.
• Background at science-focused or physical-AI company (e.g., robotics, biology).
• Deep, demonstrable expertise optimizing large-scale training and inference workloads.
• Experience with distributed training frameworks such as FSDP or DeepSpeed.
• Able to relocate and work on-site in San Francisco 5 days/week.

Responsibilities

• Design, deploy, and maintain large distributed ML training and inference clusters.
• Build scalable end-to-end pipelines for petabyte-scale datasets and training.
• Research training approaches and parallelization techniques.
• Profile and debug low-level GPU operations to optimize performance.
• Stay updated with new research and bring innovative ideas.

Skills

Distributed training

GPU optimization

Python

C++

Cloud platforms

Tools

FSDP

DeepSpeed

NVIDIA GPUs

Linux

Kubernetes

Docker

GCP

AWS

Azure

Job description

Unknown Company in San Francisco, CA seeks a lead ML infrastructure engineer to own the distributed training and inference backbone for a foundation model trained from scratch. You will stand up clusters, build data pipelines at petabyte scale, and optimize GPU performance across model scales.

You will work with FSDP/DeepSpeed, NVIDIA GPUs, Linux, Python and C++, in a distributed cloud environment across GCP/AWS/Azure, with relocation supported.