Staff Software Engineer – Inference Velocity

About the role

Anthropic’s Inference organization powers Claude for millions of users and enterprises. The Staff Software Engineer will lead the Inference Developer Productivity team, shaping toolchains, environments, and CI/CD pipelines across GPU, TPU, and Trainium platforms.

Key responsibilities

• Define technical direction and roadmap for multi‑accelerator toolchains, dev environments, and CI/CD workflows.
• Own accelerator toolchain management, keeping compilers, drivers, libraries and frameworks current, compatible and well‑tested.
• Design infrastructure for efficient accelerator usage during development, including devbox environments and validation automation.
• Instrument productivity metrics, build dashboards and alerts to surface regressions early.
• Identify and eliminate bottlenecks and toil across inference engineering workflows.
• Collaborate with Anthropic’s central Infrastructure org to align on shared productivity initiatives.
• Mentor engineers through design and code reviews, raising the technical bar.

Required profile

• 8+ years of software engineering experience with a track record as a technical lead on infrastructure or platform teams.
• Deep systems engineering background, hands‑on experience with build/test infrastructure and ML‑focused workflows.
• Demonstrated experience owning toolchains or development environments for compute‑intensive workloads.
• Strong communication skills and ability to influence technical direction without formal authority.

Required skills

• CUDA/GPU, TPU, and Trainium (AWS Neuron) accelerator ecosystems.
• ML compiler toolchains such as XLA, Triton, and NeuronX.
• CI/CD pipelines at scale, including build automation and validation.
• Kubernetes‑based development and job scheduling environments.
• Devbox or remote development environments for hardware‑dependent workflows.