Software Engineer, Infrastructure - Core Experimentation

About this role

Join OpenAI's Statsig team to build the experimentation, feature management, and analytics infrastructure powering safe product launches across OpenAI. You'll design and scale distributed systems handling low-latency configuration delivery, high-throughput event ingestion, and analytics pipelines that enable rapid, confident decision-making at company scale.

What you'll do

• Design and operate low-latency configuration delivery systems for feature flags, dynamic configs, and progressive rollouts
• Scale SDK, server-side evaluation, and control-plane systems to handle high-volume services without adding latency
• Build high-throughput data ingestion and analytics infrastructure for experimentation and product measurement workflows
• Improve performance, efficiency, reliability, and observability as OpenAI products scale globally
• Optimize query performance and data freshness for experimentation and analytics decision-making
• Lead technical initiatives and shape architecture of experimentation infrastructure used company-wide

What they're looking for

• Distributed systems design and operation
• Low-latency systems and performance optimization
• High-throughput data ingestion and analytics platforms
• Caching, concurrency, and real-time systems
• Observability, monitoring, and incident response
• Query optimization and data infrastructure
• System reliability and SLO design
• Large-scale system architecture

Benefits

• Work on mission-critical infrastructure powering OpenAI's product launches
• Based in Seattle with emphasis on in-person collaboration
• Opportunity to shape platform architecture and technical direction
• Direct impact on how company ships products safely and measures impact
• Close-knit team valuing technical depth and operational ownership

Likely interview questions

• Walk us through a large-scale distributed system you built or operated—what were the critical performance and reliability constraints?
• How would you approach designing a low-latency configuration delivery system that needs to serve millions of requests per second?