Facebook is especially interested in soliciting proposals for the wide range of AI hardware/algorithm co-design research areas, including but not limited to:
- Model compression (with particular focus on recommender systems and large embedding tables)
- Scalable communication-aware distributed training algorithms
- Processing in-memory hardware architectures for efficient and scalable machine learning
- Hardware efficiency-aware neural architecture search
- Graph-based recommender systems with implications on hardware (graph learning). Emphasis on algorithms like Deepwalk and convolutional approaches like GraphSAGE
- End-to-end hardware/software co-design automation for deep learning
- Optimizing mixed-precision linear algebra operations for AI systems
Applicants should submit a proposal detailing what contribution their research is expected to make, how the research domain will benefit from the work, project timeline and a budget overview of how the proposed funding will be used.
Proposals are highly encouraged to focus funding on project personnel, especially PhD students. Proposals from small collaborative teams, particularly with PIs bridging areas of systems and machine learning, are also encouraged. A total of six awards are available, up to to $50,000 each, depending on the specific requirements. Payment will be made to the proposer’s host university as an unrestricted gift.