MatrAIx Fellows is an applied research fellowship for PhD students and industry researchers or engineers. Each fellowship runs 1–4 months, with the possibility of extension, and requires 20–40 hours per week depending on the project. This is a hybrid program based in Mountain View, CA. Fellows work on a focused project that directly contributes to MatrAIx's core products and technology, with hands-on collaboration and mentorship from our team.
Program: 1 to 4-month fellowship, 20–40 hours/week depending on the project, with possible extension Who it's for: PhD students, and researchers or engineers in industry Focus: Personalization benchmarks and evaluation methods Location: Mountain View, CA · Hybrid
About the fellowship MatrAIx Fellows work on focused applied research projects that directly contribute to our core products and technology, with hands-on collaboration and mentorship from our team. The program suits PhD students working on agents, evaluation, user modeling or HCI who want to apply their research to real-world agent systems, and industry engineers or researchers who want to work on personalization and agent evaluation research and development. In this track, you'll develop evaluation methods to measure whether an AI agent does the right thing for a particular person. You'll work with benchmarks such as PersonalAgentEval (PAE) to find where personalization breaks down, measure it reliably, and guide improvements to MatrAIx's agent systems.
What you'll do
- Scope your project with your MatrAIx mentor
- Design new evaluation tasks or metrics for personalized, long-horizon agent behavior
- Run evaluations across models and agent systems, and analyze their failure modes
- Study how reliable graders are, including LLM-as-judge agreement with human labels
- Present progress in regular research reviews with the team
- Deliver research outcomes that directly contribute to MatrAIx's products and technology
Who we're looking for
- A current PhD student, or a researcher or engineer in industry, with a background in CS, ML or a related field
- Strong understanding of LLM agents and evaluation methodology
- Solid statistics and experimental design
- Fluent in Python and ML frameworks
- Able to commit 20–40 hours per week, depending on the project
Nice to have
- Research in personalized agents, user modeling, recommender systems or HCI
- Experience with long-horizon, browser or computer-use agent benchmarks
- Publications or open-source work in agent evaluation
Work authorization: We support CPT, OPT and STEM OPT.
Compensation & Benefits
- Stipend: $60-120 per hour
- Mentorship: Fellows receive direct mentorship from MatrAIx researchers and advisors, including 3 Nobel laureates, a Fields Medalist, and a Turing Award winner, alongside mentors with backgrounds as a former OpenAI senior director, an Anthropic researcher, Google DeepMind senior research scientist, and faculty at Stanford University, Harvard University and MIT.
- Meals: Reimbursed up to $1,500 per month
- Path to full-time: Strong fellows may be considered for full-time roles.
For any questions, please email jobs@matraix.ai