About the job
Meta is seeking AI research scientists to help us build the data foundation for Meta's most advanced Large Language Models. We're looking for researchers with LLM expertise to join us on working with data at scale and to push beyond the data ceiling.
Responsibilities
Collaborate with cross-functional teams to develop Meta’s next foundational models
Advance our understanding of data research, such as how to overcome data walls and how best to create synthetic data
Fundamentally improve our data velocity across workflows and projects by contributing to the advancement of data tooling
Architect efficient and scalable data curation systems and pipelines
Execute on high priority projects in pre-training, mid-training, or post-training data curation
Apply specialized expertise in agentic data, synthetic data, reasoning data, web parser, coding data, data scaling laws, or datamix optimization
Lead complex technical projects end-to-end
Qualifications
Minimum
Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience
PhD in Computer Science or a related technical field
2+ years of industry research experience in LLM/NLP or related AI/ML models
Experience as a formal technical lead, leading major technical initiatives with cross-functional impact, and/or influencing strategy across multiple teams
Practical experience with pre-training or mid-training data curation for large foundational models and experience working with organic, synthetic, agentic, or reasoning data for LLMs
Published research in leading peer-reviewed conferences (e.g., NeurIPS, ICML, ICLR, ACL, EMNLP) and/or demonstrated significant industry influence in the field of AI
Preferred
Experience working on frontier-quality/state-of-the-art Large Language Models
Multiple first-author publications in leading peer-reviewed conferences (e.g., NeurIPS, ICML, ICLR, ACL, EMNLP)
Hands-on experience with modeling frameworks like PyTorch
Hands-on experience on SQL and large-scale data handling, with familiarity of frameworks like Spark and Hive