SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 200,000 - 400,000 / annual
Goodfire is an AI interpretability research lab building interpretability agents and infrastructure to understand, monitor, and align AI models. The company was founded by researchers from OpenAI and Google DeepMind and has raised over $200M from leading investors.
As Compute Lead, you will own one of the company's most critical resources: compute infrastructure for model training, research, and customer workloads. You'll manage GPU allocation across competing priorities, source and secure capacity from multiple providers, and ensure infrastructure never becomes a bottleneck to research and product ambitions.
Key responsibilities include:
- Managing compute allocations across training, research, and customer workloads
- Sourcing and securing capacity: gathering availability and pricing across providers, moving quickly on optimal deals, and managing end-to-end contracting and delivery
- Owning cluster acceptance testing: defining acceptance criteria with the infrastructure team and running burn-in and performance testing before sign-off
- Monitoring cluster health post-deployment: tracking node failures, downtime, and driving fixes with providers
- Tracking utilization and quota usage by project, identifying idle capacity, and reallocating GPUs to highest-priority work
- Partnering with research and product teams to translate roadmaps into concrete compute requirements
- Building forecasting models to anticipate demand and enable long-term planning and budgeting
You'll navigate GPU scarcity and complex partner negotiations with rigor and exceptional urgency. The role requires strong judgment to balance competing demands and the conviction to make hard trade-offs. You'll report to the Head of Finance and work onsite in San Francisco 5 days per week.
The company values mission-first thinking, continuous improvement, ownership and initiative, and action-oriented execution in a high-pace environment.
Requirements: Not explicitly stated in the posting.
