Software Engineer, CI/CD & ML Infrastructure
$318k - $485k • Remote • San Francisco • FullTime
Posted 1d ago
About the job
OpenAI's Hardware organization is developing next-generation AI-native silicon and integrated systems to power frontier models. This role focuses on defining and building the CI/CD systems essential for qualifying these advanced models and the software stack that runs on OpenAI's custom AI accelerators. The work ensures that correctness, performance, and stability are measurable for every code change, from initial commits to production releases. This position involves model workloads, accelerator software, systems infrastructure, and enhancing developer productivity by establishing scalable CI/CD and GitOps practices, designing robust regression and benchmarking pipelines, and collaborating with various engineering teams to optimize feedback loops.
Responsibilities
- Define the CI strategy for frontier models on custom silicon and integrated accelerator systems.
- Build production CI/CD pipelines for regression, benchmarking, stability, and release qualification.
- Design test matrices covering models, configurations, hardware, software, and environments.
- Create reliable performance baselines, regression detection, bisect, and triage workflows.
- Establish GitOps practices for configuration management, promotion, rollback, and consistency.
- Shape monorepo architecture, dependency management, build/test boundaries, and developer workflows.
- Develop scalable orchestration, artifact management, caching, scheduling, and observability for CI.
- Partner with engineering teams to translate release risks into automated gates.
- Improve CI reliability, speed, debuggability, and cost efficiency.
Requirements
- Experience building or operating large-scale CI/CD, developer infrastructure, or production engineering systems.
- Understanding of modern software delivery practices like GitOps, release promotion, and policy-driven automation.
- Experience designing monorepos, build systems, dependency graphs, or change-impact analysis.
- Ability to design regression and performance-benchmarking systems with stable baselines and actionable failure diagnosis.
- Proficiency in Python, Go, Rust, C++, or similar languages for infrastructure and automation.
- Experience with distributed systems, schedulers, containers, clusters, or heterogeneous compute infrastructure.
- Ability to collaborate with engineers to qualify complex accelerator workloads.
- Strong focus on reliability, observability, developer experience, and fast feedback loops.