- Identify research questions that block Protocol Learning at scale, focusing on communication efficiency, convergence under churn, and heterogeneity.
- Solve complex, largely unclaimed open problems in distributed machine learning.
- Publish research findings in Tier-1 venues such as NeurIPS, ICML, and ICLR.
- Collaborate with the engineering team to ensure research methods are implemented in live training runs.
- Push research boundaries from current 8B runs toward frontier scale models.
Machine LearningPyTorchDistributed Systems