Software Engineer, Infrastructure

Thinking Machines Lab•San Francisco, CA

1d•$300,000 - $350,000•Onsite

About The Position

We're hiring a Software Engineer, Infrastructure to design and build the distributed systems that power our model training and serving platforms. You'll work on the systems underlying everything we do — from the clusters that train our frontier models with Inkling, to the multi-tenant serving infrastructure behind Tinker. This is a foundational infrastructure role at a fast-moving startup. You'll have real ownership over systems that run at large scale, and your work will directly determine how quickly our research and product teams can iterate.

Requirements

Demonstrated expertise designing and developing large-scale distributed systems
Strong proficiency in Python and Go
Experience building, deploying, and operating production infrastructure at scale
Solid grounding in distributed systems fundamentals, such as consensus, consistency, fault tolerance, and networking

Nice To Haves

Experience with ML infrastructure, such as training orchestration, job schedulers, or distributed storage and data systems
Experience operating large-scale GPU or TPU clusters
Experience with container orchestration (e.g. Kubernetes) and infrastructure-as-code
Contributions to open-source infrastructure projects
Comfortable working with high autonomy in a fast-changing, early-stage environment

Responsibilities

Design, build, and operate distributed systems that support large-scale model training and inference across thousands of accelerators
Build and maintain core infrastructure, including orchestration, scheduling, storage, and resource management systems
Improve the reliability, performance, and observability of infrastructure used across research and product teams
Partner with researchers and platform engineers to understand infrastructure needs and turn them into robust, well-abstracted systems
Debug and resolve complex distributed failures across the stack, from networking and storage to compute and scheduling
Write and maintain internal libraries and APIs, primarily in Python and Go, that other engineers build on