Research Engineer, RL Infrastructure and Reliability (Knowledge Work) (AI)

TL;DR

Research Engineer, RL Infrastructure and Reliability (Knowledge Work): Own reliability, observability, and infrastructure foundation for Knowledge Work training environments and evaluations with an accent on proactive hardening, stress-testing at scale, and high-signal metrics. Focus on building stable systems, automating operational tooling, and driving incidents to resolution for partner teams.

Location: San Francisco, CA (hybrid policy: at least 25% time in office)

Salary: $350,000 - $850,000 USD

Company

Quickly growing AI research organization building reliable, interpretable, and steerable AI systems like Claude.

What you will do

• Serve as dedicated reliability owner for Knowledge Work training environments, providing continuity and reducing operational overhead
• Own clean, canonical evaluation tools and processes, including for model releases
• Build and automate observability, dashboards, and tooling with emphasis on trusted metrics and alerts
• Proactively harden systems via load testing, fault injection, and stress testing at realistic scale
• Act as primary contact for partner teams on environment issues and drive incident resolution
• Reduce operational burden on researchers to keep them focused on research

Requirements

• Highly experienced Python engineer shipping reliable, well-instrumented production code
• Demonstrated experience operating ML or distributed systems at scale, including on-call and incident response
• Strong SRE or production-engineering mindset with SLOs, load tests, and failure injection
• Foundational ML knowledge to understand training environments, evaluations, and integrity issues
• Able to read research code and reason about evaluation integrity
• Bachelor’s degree or equivalent in relevant field

Nice to have

• 5+ years operating ML or distributed systems at scale
• Experience with RL environments, agent harnesses, or LLM evaluation frameworks
• Familiarity with reward modeling, evaluation design, or reward hacking mitigation
• Experience with observability stacks, dashboard tooling, chaos engineering, or large-scale load testing
• Background in data quality pipelines, drift detection, or evaluation curation
• Familiarity with large-scale training/inference infrastructure
• Prior role as reliability or operations owner in research team

Culture & Benefits

• Collaborative team focused on high-impact AI research as big science
• Competitive compensation, equity donation matching, generous vacation and parental leave
• Flexible working hours and lovely office space in San Francisco
• Visa sponsorship available (with reasonable effort and immigration lawyer support)
• Emphasis on diverse perspectives and representation in AI development