Dan Fu

Dan Fu is an Assistant Professor at the University of California, San Diego (UCSD) in the Computer Science and Engineering Department and a Distinguished Research Scientist at Together AI. He leads the SandyResearch Lab and is affiliated with the MLSys group, focusing on making machine learning models faster and more efficient through hardware-aware algorithms and subquadratic architectures.

• Research interests include:
  • Efficient ML architectures (Chipmunk, Hyena, Monarch Mixer)
  • Hardware-aware systems algorithms (ThunderKittens, FlashAttention)
  • Long-sequence modeling and GPU optimization

His recent work spans training-free Transformer acceleration (Chipmunk), convolutional language models (Hyena), and hardware-aware attention optimizations (FlashAttention). Projects are deployed in production at Together AI and integrated with frameworks like PyTorch.

Scientific awards include:

• NDSEG Fellowship (2025)
• Best Paper at ICML Hardware Aware Workshop (2022)
• Best Student Paper Runner Up at UAI (2022)
• Best Poster at ENLSP Workshop (NeurIPS 2023)

He contributes to open-source projects (Safari repository, FlyingSquid) and teaches courses on machine learning systems (Stanford CS 324/528, Harvard CS 61/152).