SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 320,000 - 485,000 / annual
Anthropic is seeking a Principal Engineer for the Capacity & Compute team to build and operate production systems that manage one of the largest and fastest-growing infrastructure fleets in the industry. The role spans multiple accelerator families, CPU families, and cloud providers.
You will own the data pipelines, tooling, and operational systems that enable Anthropic to plan, measure, and maximize utilization across first-party and third-party compute. This is a hands-on engineering role with a DevOps flavor, requiring you to write production-quality code daily and operate Kubernetes-native infrastructure at meaningful scale.
The work spans four interconnected areas:
**Data Platform**: Build pipelines that ingest occupancy and utilization telemetry from Kubernetes clusters, normalize billing and usage across cloud providers, and serve BigQuery tables for the organization. Focus on correctness, completeness, and latency. Consumers range from research engineers to finance to leadership, making this product work as much as engineering—defining schema contracts, ensuring data discoverability, and understanding actual consumer needs.
**Planning**: Make the state of the fleet legible and actionable in real time. Build cluster health tooling, capacity planning platforms, alerting on occupancy drops and allocation problems, and systemic fixes to scheduling and fragmentation. Balance Kubernetes operations with cross-team coordination.
**Efficiency**: Measure and improve how effectively every major workload uses its hardware. Instrument utilization across training, inference, and eval systems, build benchmarking infrastructure, establish per-config baselines, and work directly with system-owning teams to close gaps. Metrics must be defensible to the teams responsible for them.
**Attribution & Forecasting**: Connect fleet costs to business activities. Reconcile CSP billing exports against vendor telemetry and internal systems with mismatched schemas, attribute spend to workloads and teams, and turn inference demand signals and research roadmaps into defensible compute plans. Efficiency metrics must survive finance scrutiny—stripped of pure demand and unit-price effects, reproducible month-over-month, and legible to a CFO.
Key responsibilities include building the planning and allocation stack, driving efficiency programs (stranding, rightsizing, unused capacity recovery), owning attribution and forecasting, building the data platform with real ownership of completeness and SLOs, operating Kubernetes-native systems at scale, and treating outputs as products with schema contracts and consumer-focused design.
You will collaborate closely with research engineering, infrastructure, inference, and finance teams in a high-autonomy, high-ambiguity environment requiring comfort moving between data engineering, systems engineering, and observability.
**Requirements:**
- Strong track record building and operating production systems (hands-on engineering with DevOps flavor)
- Production-quality Python and SQL (idiomatic, well-tested, maintainable code)
- Deep experience with at least one major cloud provider (AWS, GCP, or Azure) and its operations
- Experience with observability tooling stack (Prometheus, PromQL, Grafana), including writing recording rules and building monitoring that engineering teams rely on
- Ability to gather own requirements and work across organizational boundaries in ambiguous environments with limited direction
- Bachelor's degree or equivalent combination of education, training, and professional experience in a field relevant to the role
**Preferred Qualifications:**
- Capacity planning, resource management, or cost attribution systems experience at a hyperscaler or large-scale ML environment
- Scheduling and packing efficiency experience or profiling-driven optimization of large distributed workloads
- Multi-cloud data ingestion experience, especially normalizing billing exports, reservation APIs, and vendor telemetry from providers with different billing arrangements
- Total cost of ownership and forecasting experience, including decomposing whether infrastructure growth is causal or correlated with business drivers
- Accelerator infrastructure familiarity (GPU metrics via DCGM, TPU utilization, Trainium power/utilization, or ML training/inference systems at hardware level)
- Experience building internal data products with self-service access, schema contracts, API serving, documentation, and discoverability
- Storage efficiency, retention, and lifecycle program experience at exabyte scale
