Network Engineer, Supercomputing

Thinking Machines Lab•San Francisco, CA

5d•$350,000 - $475,000•Onsite

About The Position

Thinking Machines Lab's mission is to empower humanity through advancing collaborative general intelligence. We're building a future where everyone has access to the knowledge and tools to make AI work for their unique needs and goals. We are scientists, engineers, and builders who’ve created some of the most widely used AI products, including ChatGPT and Character.ai, open-weights models like Mistral, as well as popular open source projects like PyTorch, OpenAI Gym, Fairseq, and Segment Anything. We're looking for a network engineer to own the lowest layers of the network stack that our large-scale training and inference depend on. A single degraded link or flapping NIC can quietly slow a long training run or take it down outright; you’ll be responsible for the reliability of the interconnect at multi-thousand-GPU scale — both the RDMA/RoCE fabric between nodes and the NVLink/NVSwitch domains within them. This is a hands-on, cross-stack role. You'll debug production collectives down to the NIC, build the instrumentation and tooling that makes the next debugging session dramatically faster, and serve as the technical point of contact who drives issues to resolution with our cloud providers' networking teams. Your goal is for our researchers to trust the fleet without worrying about the fabric underneath. Note: This is an "evergreen role" that we keep open on an on-going basis to express interest. We receive many applications, and there may not always be an immediate role that aligns perfectly with your experience and skills. Still, we encourage you to apply. We continuously review applications and reach out to applicants as new opportunities open. You are welcome to reapply if you get more experience, but please avoid applying more than once every 6 months. You may also find that we put up postings for singular roles for separate, project or team specific needs. In those cases, you're welcome to apply directly in addition to an evergreen role.

Requirements

Bachelor’s degree or equivalent experience in computer science, engineering, or similar.
Proficiency in at least one backend language (we use Python or Rust).
Experience operating large‑scale clusters and container orchestration systems (e.g. Kubernetes or Slurm).
Comfort operating across the stack and owning projects end-to-end.
Thrive in a highly collaborative environment involving many, different cross-functional partners and subject matter experts.
A bias for action with a mindset to take initiative to work across different stacks and different teams where you spot the opportunity to make sure something ships.

Nice To Haves

Fluency with host-level debugging tools on Linux.
Strong communication skills, internally and with cloud providers.
Familiarity with cloud network primitives across at least two cloud providers.
Hands-on experience with NVLink / NVSwitch, fabric manager, and IMEX.
Statistical rigor in reliability reasoning — comfort reasoning about failure and error rates, distributions, and base rates, and the judgment to separate signal from noise when characterizing a large fabric.
A track record of writing tooling that made the next debugging session meaningfully faster.
Familiarity with CUDA/NCCL and performance profiling for distributed training and inference.
Understanding of deep learning frameworks and their underlying system architectures.

Responsibilities

Reason about and validate GPU network fabric design across our deployments.
Debug RDMA / RoCEv2 across different NIC vendors. Diagnose collective failures of production NCCL, PFC/ECN tuning, and congestion control behavior.
Own NVLink / NVSwitch interconnect — including fabric manager and IMEX health, link and lane errors, and how the GPU fabric interacts with collectives.
Build host-level network instrumentation and use Linux tooling to build dashboards and alerts, not just the bug report.
Navigate cross-cloud fabric quirks across providers and triage across the NIC, driver, kernel, switch, and workload boundaries.
Drive escalations with cloud-provider networking teams, owning issues end-to-end until they're resolved.