ML Infrastructure Engineer (AI Document Processing)

A rapidly growing AI startup that turns unstructured documents into usable data needs an engineer to run and scale how its models are trained and served. This is an individual-contributor position with a lot of leverage, working beside researchers so that infrastructure never limits what ships.

What you’ll own

• Maintain and improve model serving, increasing inference speed, uptime and dependability for customers
• Set up and refine training environments for models from a few hundred million to about thirty billion parameters on small multi-node setups
• Develop observability, logging and alerting across the entire ML stack
• Create data pipelines and internal tools that carry researchers from prototype to production faster
• Design systems that route inference across several cloud providers while balancing accuracy, latency and spend

Profile

• Three or more years in ML infrastructure with a focus on serving and training
• Strong Python and systems engineering skills, including Kubernetes
• Deep knowledge of GPU bottlenecks, serving optimization and monitoring
• A generalist who understands how ML models behave, not someone narrowly specialized in kernels or in massive-scale training alone
• Practical experience with small-node training and serving is a plus
• Comfortable working closely with researchers at a startup pace, and native to AI tooling

On-site in the San Francisco Bay Area. Visa transfers considered.