AI Engineer, RL

Taste Labs is building systems to tackle subjective “taste” challenges with practical evaluation methods and scalable reinforcement learning infrastructure. This role focuses on creating grading approaches, tasks, and environments that let models learn and improve without humans in the loop, while partnering closely with internal research teams and frontier labs.

What you’ll work on

• Explore and compare grading methods and rubric designs for subjective domains
• Design tasks that capture elements of “taste” and define task capabilities
• Build agent harnesses and context layers to support RL workflows
• Contribute to scalable RL infrastructure that can support training and evaluation at scale
• Collaborate with internal research teams on training pipelines
• Work with top frontier labs on crafting environments that improve frontier models

What you bring

• Experience building evals, RL environments, and ML or post-training, along with strong backend engineering skills
• A preference for ambiguous, hard, and creative problems, paired with a drive to make subjective domains verifiable
• Team player mindset and startup-style execution: moving fast, adapting, operating without strict boundaries, and taking ownership of outcomes

Bonus points

• Open source contributions or personal projects that demonstrate building driven by curiosity
• Background at creative companies (Figma, Notion, Canva, Adobe, Runway, etc.) or companies with strong index building/crawling (Firecrawl, Brave, Luma, Pika) or data-focused companies (Mercor, Surge, etc.)

Location and compensation

San Francisco, CA (onsite). Salary range: USD 175,000 - 275,000 per year.