Values

The ad says“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.”
The ad says“Experience with cloud-native infrastructure across AWS/GCP/Azure, Infrastructure as Code, and CI/CD.”
The ad says“Experience with cloud-native infrastructure across AWS/GCP/Azure, Infrastructure as Code, and CI/CD.”
The ad says“Experience with cloud-native infrastructure across AWS/GCP/Azure, Infrastructure as Code, and CI/CD.”
The ad says“Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent.”
The ad says“Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent.”
The ad says“Experience with cloud-native infrastructure across AWS/GCP/Azure, Infrastructure as Code, and CI/CD.”
The ad says“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).”
The ad says“Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.”
The ad says“Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent.”
The ad says“Experience with cloud-native infrastructure across AWS/GCP/Azure, Infrastructure as Code, and CI/CD.”
The ad says“Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent.”
The ad says“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.”
The ad says“Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar).”
The ad says“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).”
The ad says“Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.”
The ad says“Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.”
The ad says“Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.”
The ad says“Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.”
The ad says“Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.”
The ad says“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).”
The ad says“Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar).”
The ad says“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.”
The ad says“Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar).”
The ad says“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).”
The ad says“Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar).”
The ad says“Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar).”
The ad says“Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent.”

About Scale AI

Scale Frontier Data is the organization behind the training and evaluation data that frontier labs depend on. We build the systems, tooling, and expert workflows that turn hard human expertise into signals that models can learn from, across reasoning, coding, agentic tool use, and domain expertise.

Reinforcement learning environments are now the center of gravity for that work: the difference between a model that demos well and a model that reliably completes long-horizon work is almost always the quality of the environments and reward signals it was trained against.

Responsibilities

As a Staff Software Engineer, RL Environments, you'll own the technical foundation for 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.

Required Qualifications

8+ years of software engineering experience with strong fundamentals in distributed systems, system design, data structures, and algorithms.
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).
Deep experience with containerization and sandboxed execution, including Docker, VMs, gVisor/Firecracker, Kubernetes, or equivalent.
Experience building or operating high-throughput backend systems: orchestration, job scheduling, queuing, and large-scale data pipelines.
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.
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.

Preferred Qualifications

RL & Post-Training

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).
Familiarity with post-training methods: RLHF, RLAIF, RLVR, GRPO/PPO-family algorithms, rejection sampling, reward modeling, and the practical failure modes of each.
Experience designing verifiable reward signals, and firsthand experience with reward hacking and how to defend against it.
Experience with RL training or serving stacks (verl, TRL, Ray, vLLM, SGLang, or similar).

Systems & Infrastructure

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.
Strong observability instincts: tracing, structured logging, and metrics for systems whose failure modes are statistical rather than binary.
Experience building internal tools that non-engineers rely on daily, especially data-dense review and annotation interfaces.

Ways of Working

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.

Compensation packages at Scale for eligible roles include base salary, equity, and benefits. The range displayed on each job posting reflects the minimum and maximum target for new hire salaries for the position and may be inclusive of several career levels at Scale; it will be determined during the interview process based on work location and additional factors, including job-related skills, experience, qualifications, interview performance, and relevant education or training. Scale employees in eligible roles are also granted equity based compensation, subject to Board of Director approval. Your recruiter can share more about the specific salary range for your preferred location during the hiring process, and confirm whether the hired role will be eligible for equity grant. You'll also receive benefits including, but not limited to: comprehensive health, dental and vision coverage, retirement benefits, a learning and development stipend, and generous PTO. Additionally, this role may be eligible for additional benefits such as a commuter stipend.

Please reference the job posting's subtitle for where this position will be located. For pay transparency purposes, the base salary range for this full-time position in the locations of San Francisco, New York, Seattle is:

$252,000—$315,000 USD

PLEASE NOTE: Our policy requires a 90-day waiting period before reconsidering candidates for the same role. This allows us to ensure a fair and thorough evaluation of all applicants.

About Us:

At Scale, our mission is to develop reliable AI systems for the world's most important decisions. Our products provide the high-quality data and full-stack technologies that power the world's leading models, and help enterprises and governments build, deploy, and oversee AI applications that deliver real impact. We work closely with industry leaders like Meta, Ernst & Young, Mayo Clinic, Time Inc., the Government of Qatar, and U.S. government agencies including the Army and Air Force. We are expanding our team to accelerate the development of AI applications.

We believe that everyone should be able to bring their whole selves to work, which is why we are proud to be an inclusive and equal opportunity workplace. We are committed to equal employment opportunity regardless of race, color, ancestry, religion, sex, national origin, sexual orientation, age, citizenship, marital status, disability status, gender identity or Veteran status.

We are committed to working with and providing reasonable accommodations to applicants with physical and mental disabilities. If you need assistance and/or a reasonable accommodation in the application or recruiting process due to a disability, please contact us at accommodations@scale.com. Please see the United States Department of Labor's Know Your Rights poster for additional information.

We comply with the United States Department of Labor's Pay Transparency provision.

PLEASE NOTE: We collect, retain and use personal data for our professional business purposes, including notifying you of job opportunities that may be of interest and sharing with our affiliates. We limit the personal data we collect to that which we believe is appropriate and necessary to manage applicants’ needs, provide our services, and comply with applicable laws. Any information we collect in connection with your application will be treated in accordance with our internal policies and programs designed to protect personal data. Please see our privacy policy for additional information.