Drimva

Meta

Research Scientist Intern, MSL Agents (PhD)

Menlo Park, CA · Internship · Posted Oct 22, 2026

Education
Ph.D. in Computer Science, Artificial Intelligence, Generative AI, or a relevant technical field; currently holds or is pursuing the degree.
Visa sponsorship
Must obtain and maintain work authorization in the country of employment.

About the job

Conduct research and develop, evaluate, and deploy post-trained LLM systems and long-horizon agents for Meta products.

What you'll do

  • Collaborate with researchers and cross-functional partners, and communicate research plans, progress, and results.
  • Build training data, design and run product-relevant evaluations, and execute and analyze large-scale post-training runs.
  • Advance agent capabilities and research training, evaluation, synthetic-data generation, and data-curation strategies.

What they ask for

  • Currently has or is pursuing a Ph.D. in Computer Science, Artificial Intelligence, Generative AI, or a relevant technical field.
  • Experience with Python, C++, C, Java, or related programming languages, and with machine learning or deep learning systems.
  • Intent to return to the degree program after the internship and work authorization in the employment country throughout employment.
  • Demonstrated research results, such as grants, fellowships, patents, or first-authored publications at leading workshops or conferences.

Nice to have

  • Experience advancing AI techniques, including contributions to open-source computer vision libraries or frameworks.
  • Experience with deep learning frameworks, ML experiments, large-scale data analysis, and quantitative research.

Pay

$7,650/month to $12,667/month (USD; base salary)

Tools

Python, C++, C, Java, NeurIPS, ICML, ICLR, CVPR

AI focus

Research and post-train LLMs for long-horizon agents, tool use, full-stack coding, search, planning, and personalization; develop training data, evaluations, and synthetic-data strategies.

Model-written summary · the original posting is authoritative

First seen Oct 23, 2026 · Last verified Nov 5, 2026 (UTC)