About the company Our client is a software company. The role Raydar is recruiting for this role on behalf of our client. Design and build realistic simulated work environments, scoring systems and evaluation tasks used to train and assess AI models. The work covers generating detailed synthetic workplace data, hardening graders against shortcuts, and turning real-world feedback into new test cases. Candidates with a background in model post-training or reinforcement learning are a strong fit for this early-career role. What you'll do
- Build simulated environments with generated records, files and communications that behave like real workplaces.
- Design graders and scoring functions that resist exploitation by models.
- Create benchmarks whose rankings track real-world performance.
- Convert user feedback and corrections into evaluation tasks.
- Take over environment and benchmark work currently led by the founders.
Requirements
What we're looking for
- 0 to 3 years of experience.
- A degree from a top-30 school.
- A background in post-training or reinforcement learning, from research or industry.
- A good sense of what makes a simulated environment rich, detailed and believable.
- The ability to spot how a model might game an environment or grader, and to fix it.
- Strong enthusiasm for using AI tools to work more effectively.
Benefits
Compensation and benefits
- Base salary: USD 150,000 to 300,000 per year
- Equity Location and work model
- San Francisco, CA, United States
- On-site
- Full-time