About the role
Trata is hiring a Research Scientist Intern for the summer. If post-grad, contract work is also an option. We are creating very unique benchmarks and evaluations. You will design datasets and evaluation rubrics that directly influence how frontier models learn. You will have the opportunity to co-write a paper that is largely the first of its kind. Your output will feed directly into model training runs at scale. This role requires facetime in NYC and SF (we pay for travel) but can include a remote portion for the right applicant.
What You'll Do
• Design data slices that expose meaningful model failure modes
• Build and refine evaluation rubrics and reward signals for RLHF and RLVR training pipelines
• Develop quantitative frameworks for measuring dataset quality, diversity, and downstream impact on alignment and capability
• Partner with lab research teams to translate training objectives into concrete data and evaluation specifications
What We’re Looking For
• Deep curiosity and motivation around how data structure, selection, and quality drive model behavior
• Ability to design lightweight experiments, move fast, and extract actionable insights from messy results
• A bias toward building over theorizing
• Prior experience at an AI safety org, benchmarking org, or RL environment company is a strong plus
Strong performers will be considered for a full-time offer.
To Apply
Send a brief note and your CV Links to evals you've built, experiments you've run, or writing on post-training are strongly encouraged.
About Trata:
