Staff Software Engineer (RL Environments)

• How Scale builds, runs, verifies, and delivers RL environments at scale
• An RL environment is a real piece of software: a containerized world with real dependencies, real state, real tools, and a grader that has to be correct even when the agent is creative about breaking it. Building one is a full-stack engineering problem. Building thousands of them reproducibly, cheaply, with trustworthy reward signals and throughput measured in millions of rollouts is a systems problem that very few people have solved
• You’ll work on both. You’ll design the platform: sandboxed execution, environment packaging and versioning, rollout orchestration, trajectory capture, verifier frameworks, and the authoring surfaces that let engineers and domain experts produce environments without reinventing infrastructure each time. And you’ll go deep on the environments themselves by instrumenting real applications, designing task suites that expose specific capability gaps, and building graders that hold up under adversarial optimization
• This is a hands-on engineering role. You’ll set technical direction across multiple teams, and you’ll still be the person who writes the hard part

Benefits

• Health & Wellbeing: Our holistic approach to supporting Scaliens includes comprehensive health coverage, dental and vision insurance, mental healthcare services, and more. PTO policies and accommodating schedules ensure you’ll get time off when you need it to relax and recharge. Note that our offerings may vary by region as we strive to respond to the unique needs of Scaliens around the globe.
• Personal & Career Growth: Continuously learn and grow through annual learning & development stipend, attending leadership breakfasts, manager training, speaker series, and joining an ERG.
• Building Scale Community: We welcome guests to our offices, and you can expect to see Scalien families and friends around. Join local happy hours, and accept invites to game nights, book clubs, and many other employee-led community events.
• Parental Support: Balancing work and family is essential, and Scale understands the importance of having adequate leave policies in place to promote a healthy home and work life.- Hands-on experience building with LLMs including agent loops, tool calling, MCP, or eval harnesses, and enough intuition about model behavior to reason about what a training signal actually teaches
• Experience building or operating high-throughput backend systems: orchestration, job scheduling, queuing, and large-scale data pipelines
• Demonstrated ability to own ambiguous, undefined problems end to end and drive them to a shipped system
• Excellent written and verbal communication; ability to align engineers, researchers, and non-engineering partners on a technical direction
• Strong Python skills and a track record of shipping production software; comfort in at least one other part of the stack (TypeScript/React, Go, Rust, or similar)
• 8+ years of software engineering experience with strong fundamentals in distributed systems, system design, data structures, and algorithms
• Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent
• Experience with high-scale sandbox or code-execution infrastructure, remote development environments, or CI systems
• Experience with cloud-native infrastructure across AWS/GCP/Azure, Infrastructure as Code, and CI/CD
• Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar)
• Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each
• Experience building internal tools that non-engineers rely on daily, especially data-dense review and annotation interfaces
• Experience designing verifiable reward signals, and firsthand experience with reward hacking and how to defend against it
• Strong observability instincts: tracing, structured logging, and metrics for systems whose failure modes are statistical rather than binary
• Direct experience building RL environments, agentic benchmarks, or eval harnesses (SWE-bench-style task suites, terminal or browser environments, tool-use benchmarks, or in-house equivalents)
• Experience in a research-adjacent engineering role, translating research goals into production systems
• Experience working directly with sophisticated external technical customers
• Prior technical leadership at staff level or above in a fast-moving, ambiguous environment