Surge AI jobs, judged for you, free

What they do

Surge AI gives AI labs the human data that trains and tests their models. Its platform pairs expert human workers with tools for data labeling, evaluations, and RL environments. RL is reinforcement learning, where a model learns from rewards. It works with Anthropic, Google, Microsoft, and Meta. It made a profit from the start without venture funding.

What they value

• AI with the richness of human intelligence: curious, witty, creative, and wise.
• Rich human data over scale alone, and AI that is useful, not tuned for clicks.
• Full ownership of work, from idea to production.
• Autonomy, quick shipping, and a high bar for quality.
• AI ethics: models that are safer and more useful for society.
• Taste for good engineering: correctness, code quality, and real team practice.

Where they hire

All 35 open roles are full time and remote in the United States. The security engineering role says the team is based in San Francisco and New York. It asks people in other places to travel at times to meet the engineers. Postings do not mention visa sponsorship, and they give no pay ranges.

Teams and roles

Engineering postings call Surge a small team at the center of the AI field. Fifteen of the 35 roles are in Field Engineering and Operations, which runs data projects for AI labs. The rest are in platform engineering, research and data science, legal, corporate operations, partnerships, marketing, and finance.

• Staff Fullstack Engineer and Staff Product Engineer: build the core platform for data collection, labeling, and evaluation.
• Research Engineer and Software Engineer, Coding Evaluation and Training Data: design coding tasks, RL environments, and grading rules that teach models to code.
• Software Engineer, Security: owns security across product, infrastructure, and IT, and fixes problems directly.
• RL Environments Architect: designs the simulated worlds where AI agents learn.
• Forward Deployed Researcher and Research Scientist: work with AI labs to study model behavior and design training datasets.
• Lead Adversarial Engineer: runs red-team campaigns against frontier models. A red team attacks a system to find its weak points.
• AI Operations Engineer: an early-career role that builds data pipelines and internal tools.

What they look for

The staff engineering and security roles ask for seven or more years of building production software. The staff roles ask for React and TypeScript, and Python or Ruby on Rails. All three also ask for early-stage startup experience and a CS degree from a top school or an equal STEM background. The coding evaluation roles ask for two to six or more years of software engineering. They ask for skill with containers and automated test setups.

Research roles ask for curiosity about how models reason and fail, and for careful experiment design. The Human Data Architect role asks for knowledge of ML post-training, such as SFT and RLHF, two ways to tune a trained model. Program and operations roles ask for skill with ambiguous problems and for influence without authority. Many roles ask for excellent communication. Most program and operations roles ask for an interest in AI, but they do not need direct AI or ML experience.