Staff+ Software Engineer, Inference Runtime
Technical lead for the shared, accelerator-agnostic inference runtime serving Claude. Owns architecture, performance, and validation for GPU/TPU/Trainium platforms in a high-scale distributed systems environment.
About the job
Key Responsibilities
- Set technical direction for the team, owning the architecture and roadmap for the shared runtime of the inference serving stack
- Own and evolve the accelerator-agnostic runtime itself – its interfaces, internal boundaries, and build structure – including hands-on work in a performance-sensitive Rust and Python codebase
- Keep the platform's expansion cost low by ensuring new models and deployment targets pay only for their own specialization, and edge cases stitch back into the core easily
- Drive efficient accelerator usage – utilization, scheduling, memory management – across GPU, TPU, and Trainium
- Build the runtime's validation surface around partitioned builds, change-scoped testing, and canary/shadow/rollback as first-class mechanisms
- Act as a technical counterpart to Anthropic's central Infrastructure org on the compilers, build systems, and toolchains the runtime depends on, contributing Inference's performance and correctness requirements, and making the call on build vs. adopt
- Mentor engineers on the team through design review, code review, and direct collaboration, raising the technical bar without owning headcount
Minimum Qualifications
- Deep background in systems engineering or ML infrastructure, with the ability to go hands-on with performance profiling, latency and throughput optimization, and systems debugging at scale
- Real depth in at least one accelerator ecosystem (CUDA/GPU, TPU, or Trainium/AWS Neuron) and genuine appetite to keep the runtime agnostic across all of them
- Have significant software engineering experience, with a strong background in high-performance, large-scale distributed systems serving millions of users
- A track record of defining and using engineering metrics to drive improvement: you've set SLOs on platform surfaces, and driven escape rates, release times, latency, or throughput in a measurable direction
- Experience driving technical alignment across organizational boundaries, advocating for your team's needs while contributing to shared infrastructure
- Strong written and verbal communication, and the ability to influence technical direction without formal authority
Preferred Qualifications
- 8+ years of software engineering experience, with significant time as the technical lead or anchor on a platform, inference runtime, or ML infrastructure team
- Experience with ML compiler toolchains (XLA, Triton, NeuronX) or accelerator driver/firmware management at scale
- Background operating production as a validation surface at scale: shadow traffic, canary populations, automated baseline comparison, fast rollback
- Experience with deterministic or simulation-based testing for hardware-dependent systems
- Experience with CI/CD systems at scale, particularly for workloads involving accelerator hardware
- Familiarity with Kubernetes-based development and job scheduling environments
- Prior tech lead experience on a developer productivity or platform engineering team at a fast-growing AI/ML company
Skills
Rust, Python, CUDA, GPU, Tpu, Aws Neuron, Xla, Triton, Kubernetes, ML Infrastructure
Similar jobs
ML Engineering jobs
Anthropic
San Francisco, CA
Build and operate scalable ML inference infrastructure for Claude’s safety systems, translating safety research into reliable production deployments. The role requires deep production ML infrastructure experience, distributed systems expertise, and proficiency with Python and modern ML frameworks.
Garner Health
New York, NY
Leads end-to-end development of production algorithmic systems for healthcare, spanning machine learning, optimization, and LLM applications. The player-coach role requires 6+ years of industry experience, strong problem-solving and metrics judgment, and technical leadership of a small team.
Garner Health
New York, NY
Leads the reliability, architecture, deployment automation, and monitoring of production machine learning systems. Requires 7+ years of software engineering experience, deep MLOps platform expertise, and strong Kubernetes, cloud, infrastructure-as-code, and observability fundamentals.
United States
Leads technical strategy for Reddit’s Ads ML Platform, improving feature development, training-data generation, experimentation, and the path to production ML serving. The role requires 8+ years in infrastructure or distributed systems, production ML platform experience, and strong cross-team technical leadership.
Square
San Francisco, CA
Build and operate production machine learning systems for ranking, retrieval, recommendations, personalization, and customer intelligence. The role requires 12+ years of production software and ML experience, strong expertise in intelligent systems, and sound judgment around trustworthy customer-impacting signals.