Staff + Senior Software Engineer, Inference Deployment

Anthropic; San Francisco (CA); On-site; USD 120,000 - 160,000; Full time

Job summary

Deepstreamtech is seeking a Software Engineer for the Launch Engineering team in San Francisco, CA. The role focuses on designing and building deployment infrastructure for inference models, optimizing scheduling against constrained resources like GPUs and TPUs. Ideal candidates should have extensive experience in creating deployment systems at scale, strong software engineering skills, and proficiency in Kubernetes deployments. Deepstreamtech is committed to making inference deployment seamless and unattended while ensuring high performance and reliability.

Qualifications

• 5+ years of experience building deployment infrastructure at scale.
• Experience designing systems managing complex state machines and multi-stage pipelines.
• Strong communication and ability to work closely with various teams.

Responsibilities

• Design and build deployment infrastructure for inference code.
• Optimize deployment scheduling to manage constrained resources.
• Ensure continuous and unattended deployment across fleets.

Skills

Experience building deployment infrastructure

Strong software engineering skills

Proficiency with Kubernetes-based deployments

Communication skills

Experience with resource-constrained scheduling

Experience in Python or Rust

Job description

Requirements

• 5+ years of experience building deployment, release, or delivery infrastructure at scale
• Strong software engineering skills with experience designing systems that manage complex state machines and multi-stage pipelines
• Experience with deployment systems where resource constraints shape the design — whether that's fleet capacity, network bandwidth, hardware availability, or coordinated rollout windows
• A track record of building automation that measurably improves deployment velocity and reliability
• Proficiency with Kubernetes-based deployments, rolling update mechanics, and container orchestration
• Comfort working across the stack — from backend services and databases to CLI tools and web UIs
• Strong communication skills and the ability to work closely with oncall engineers, model teams, and infrastructure partners
• (Desirable) Experience with ML inference or training infrastructure deployment, particularly across multiple accelerator types (GPU, TPU, Trainium)
• (Desirable) Background in capacity planning or resource-constrained scheduling (e.g., bin-packing, fleet management, job scheduling with hardware affinity)
• (Desirable) Experience with progressive delivery in systems with long validation cycles: canary/soak testing, blue-green deployments, traffic shifting, automated rollback
• (Desirable) Experience at companies with large-scale release engineering challenges (mobile release trains, monorepo deployments, multi-datacenter rollouts)
• (Desirable) Experience with Python and/or Rust in production systems

What the job involves

• Our mandate is to make inference deployment boring and unattended
• Anthropic serves Claude to millions of users across GPUs, TPUs, and Trainium — and every model update must reach production safely, quickly, and without disrupting service. We're building the systems that make inference deployment continuous and unattended
• As a Software Engineer on the Launch Engineering team, you'll design and build the deployment infrastructure that moves inference code from merge to production
• This is a resource-constrained optimization problem at its core: validation and deployment consume the same accelerator chips that serve customer traffic — your deploys compete with live user requests for the same hardware
• Every model brings different fleet sizes, startup times, and correctness requirements, so the system must adapt continuously. You'll build systems that navigate these constraints — orchestrating validation, scheduling deployments intelligently, and driving down cycle time from merge to production
• If you've built deployment systems at scale and gravitate toward the hardest problems at the intersection of automation and resource management, this team will give you an outsized scope to work on them
• Own deployment orchestration that continuously moves validated inference builds into production across GPU, TPU, and Trainium fleets, unattended under normal conditions
• Improve capacity-aware deployment scheduling to maximize deployment throughput against constrained accelerator budgets and variable fleet sizes
• Extend deployment observability — dashboards and tooling that answer "what code is running in production," "where is my commit," and "what validation passed for this deploy"
• Drive down cycle time from code merge to production with pipeline architectures that minimize serial dependencies and maximize parallelism
• Optimize fleet rollout strategies for large-scale deployments across thousands of GPU, TPU, and Trainium chips, minimizing disruption to serving capacity
• Evolve self-service model onboarding so that new models can be added to the continuous deployment pipeline without Launch Engineering involvement
• Partner across the Inference organization with teams owning validation, autoscaling, and model routing to integrate deployment automation with their systems