Job Description

This role is within the Agent Data and Optimization team. The team is the Applied AI org’s data and research environment engine: we turn world-class technical expertise into the training data, reinforcement learning (RL) environments, evaluations, and feedback loops that make Meta’s models and the products built on them the best in the world. Our research and agentic work directly improves Muse family of models and powers products such as Muse Code and the Muse assistant. This unique data and research infrastructure is a key differentiator for Meta, complementing our investments in compute and model architecture.

Why this is exciting: You’ll work directly on frontier AI models, not adjacent to them. We’re a significant, well-funded organization with the ethos of an AI startup: new, highly entrepreneurial, flexible, and already delivering measurable impact. Research and RL environments from our team went directly into the latest Muse release and helped the Muse family of models compete at the frontier.

As a Research Scientist, you’ll independently own ambitious applied research from question to measured result. You’ll identify capability gaps, formulate hypotheses, develop methods and benchmarks, build experimental environments and high-quality RL environments, and analyze agent behavior to distinguish genuine progress from shortcuts. The work centers on advancing AI through:

• Original research that develops agents’ ability to reason, experiment and solve complex technical problems.
• Evaluation and benchmarks that measure meaningful progress and generalization.
• High-quality RL environments grounded in expert knowledge.

Why Join Us

• Shape what frontier models can do

Your expertise helps identify the capabilities to develop and the evidence used to measure progress. Your work will directly improve the family of Muse models and push the boundaries of what frontier AI can do.

• Own the research from idea to results

Take a research question from definition through implementation, experiments and measured results. You’ll combine research judgment with hands-on engineering and help set the technical direction of your work.

• Work on open research questions

Investigate how agents learn from feedback, make progress across successive experiments and apply what they learn to new problems. Develop methods that can distinguish genuine improvement from errors and shortcuts.

• Collaborate across research and engineering

Work with other domain experts, engineers and researchers across Applied AI. Build on shared knowledge, challenge assumptions and turn promising ideas into reproducible methods and systems.

Depending on your interests and strengths, your work may focus on model learning, evaluation methods, research environments, or data and experimental infrastructure. The role combines deep technical judgment with implementing and testing ideas.

Responsibilities

Identify capability gaps and formulate research questions that advance AI agents’ reasoning and problem-solving abilities.

• Develop research methods, benchmarks and experimental environments with reliable, reproducible measurements.
• Create and curate high-quality data, and build pipelines that support model training and evaluation.
• Analyze agent behavior and design feedback that rewards useful progress and tests generalization to new problems.
• Lead research from hypothesis through implementation and measured results, collaborating with Meta’s model development and research teams.

Qualifications

A PhD in Computer Science, AI, relevant technical field, or equivalent practical experience

• 3+ Years Experience leading major research initiatives or influencing technical strategy across teams
• Demonstrated ability to build reliable technical systems and evaluate their results
• A strong track record of first-author publication on AI-related topics with experience presenting the work
• Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience