Global Capacity Manager - TPU Focus

Baseten
San Francisco
Workplace: HybridFull timeUSD 185,000 - 250,000 annuallyFunction: DevOps, Cloud & InfrastructureExperience: 5+ yearsSkills: ["Entrepreneurial mindset","Collaboration","High tenacity","Financial modeling","Stakeholder partnership"]

Own end-to-end capacity management for Baseten’s non-NVIDIA accelerator fleet, with a focus on Google Cloud TPUs. Architect, secure, and optimize large-scale TPU pod allocations, build automation for reliable uptime across multi-cloud environments, and orchestrate migrations and deployment drains using Kubernetes. Lead incident response during capacity crunches, partner with Model Performance, and create ROI models comparing TPU, GPU, and other accelerators to scale profitably.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
Baseten
Baseten
13 hours ago

Global Capacity Manager - TPU Focus

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 1 hour agoStatus: Live

Job Summary

Own end-to-end capacity management for Baseten’s non-NVIDIA accelerator fleet, with a focus on Google Cloud TPUs. Architect, secure, and optimize large-scale TPU pod allocations, build automation for reliable uptime across multi-cloud environments, and orchestrate migrations and deployment drains using Kubernetes. Lead incident response during capacity crunches, partner with Model Performance, and create ROI models comparing TPU, GPU, and other accelerators to scale profitably.
Location: San Francisco
Workplace: Hybrid
Employment Type: Full time
Job Function: DevOps, Cloud & Infrastructure

Key Responsibilities

  • •Lead specialized TPU pod fleets across the acquisition, allocation, and maintenance lifecycle.
  • •Execute advanced orchestration, including workload migrations and sticky deployment drains across TPU topologies with strict scheduling/compliance requirements.
  • •Design and implement the next version of Baseten’s capacity management system to scale TPU volume alongside the GPU fleet.
  • •Develop financial models and ROI comparisons across TPU, GPU, and other accelerators to ensure profitable growth.
  • •Collaborate with MP, SRE, Infra, and FDE teams and lead rapid capacity-crunch incident response by reallocating and re-coordinating TPU workloads.

Pay and Benefits

Salary: USD 185,000 - 250,000 annually
Equity and Bonus:Equity
Perks:Health InsuranceDentalVision401kPaid ParentalFertility Stipend

Key Requirements

  • •Bachelor’s, Master’s, or Ph.D. in Computer Science, Engineering, Mathematics, or a related field.
  • •5+ years of professional experience in a high-growth environment, preferably at a hyperscaler or specialized accelerator provider.
  • •Hands-on experience with Google Cloud TPUs, including pod slicing, ICI topology, JAX/XLA, and TPU scheduling/fault handling.
  • •Deep expertise in Kubernetes, including taints, cordons, node draining, and custom operators.
  • •Production experience with Go or Python and strong ability to model capacity reliability vs. cost trade-offs.
Experience:5+ yearsHigh-growth environmentHyperscalerMulti-cloudMachine learning infrastructureAccelerators
Education:
Skills:Entrepreneurial mindsetCollaborationHigh tenacityFinancial modelingStakeholder partnership
Tech Stack:Google CloudTPUsTrilliumKubernetesPod slicingICI topologyJAXXLAGoPythonMulti-cloud capacity managementROI modeling

Company Brief

Baseten
Baseten provides an inference-first ML infrastructure platform that lets engineering and ML teams deploy, serve, and scale machine-learning models with optimized performance, autoscaling, and GPU-backed hosting for production AI applications. ([crunchbase.com](https://www.crunchbase.com/organization/baseten?utm_source=openai))
Industry: AI & Machine Learning
Company Size: Medium (51 to 250 employees)
Growth: Scaleup
Valuation: Unicorn (USD 1B+)
Funding: Series E+
Headquarters: San Francisco, United States
Founded: 2019
Glassdoor
Glassdoor: 5.0
WebsiteLinkedInGlassdoor