Research Scientist Intern, Core ML for Recsys (PhD)

Meta; Bellevue (WA); On-site; USD 85,000 - 141,000; Full time

Job summary

Meta seeks a Research Scientist Intern for the Meta Recommendation Systems core team to advance AI/ML research across multi-modal signals and large-scale models. You will build frameworks and train models deployed with product teams to improve recommendation relevance at scale.

The role emphasizes collaboration with researchers and cross-functional partners, with opportunities to publish papers and contribute to open-source software. A PhD-background is preferred but not required for all tracks.

Qualifications

• Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment.
• Experience with Python, with experience in machine/deep learning libraries such as Pytorch and Transformers.
• Familiarity with AI/ML modeling and algorithmic techniques (e.g., multimodal LLM, RAG, LSTM, GRU, Transformers, RL) and emergent technologies such as agentic AI.

Responsibilities

• Initiate and lead efforts towards long-term ambitious research goals in recommendation systems and multimodal analysis.
• Conduct original research that can be applied to Meta product development and publish/open source software where appropriate.
• Design, train and support AI/ML libraries and models for internal Meta use.
• Collaborate with researchers and cross-functional partners, communicating plans, progress, and results.

Skills

Python

Pytorch

Transformers

Multimodal ML

Job description

Meta seeks a Research Scientist Intern for the Meta Recommendation Systems core team to advance AI/ML research across multi-modal signals and large-scale models. You will build frameworks and train models deployed with product teams to improve recommendation relevance at scale.

The role emphasizes collaboration with researchers and cross-functional partners, with opportunities to publish papers and contribute to open-source software. A PhD-background is preferred but not required for all tracks.