Beidi Chen

Beidi Chen is an Assistant Professor in the Department of Electrical and Computer Engineering at Carnegie Mellon University (CMU), with a visiting research role at FAIR/Meta. She holds a Ph.D. in Computer Science from Rice University (2020) and a B.S. in Electrical Engineering and Computer Science from UC Berkeley (2015). Her research focuses on optimizing machine learning algorithms and models for modern hardware, emphasizing efficiency in large-scale systems. She has been recognized as a Rising Star in EECS by MIT (2019) and UIUC (2021), and her work has received multiple best-paper awards, including runner-up at ICML 2022 and awards at IISA 2018 and USENIX LISA 2014.

• Education:
  • Ph.D., Computer Science, Rice University (2020)
  • B.S., Electrical Engineering & Computer Science, UC Berkeley (2015)

Her research interests span machine learning systems, algorithm-system co-design, and large language models (LLMs). She explores techniques like structured sparsity, efficient inference, and hardware acceleration to enhance the scalability and performance of ML models. Recent work includes optimizing LLM context lengths, speculative decoding, and memory-efficient training methods.

Her publications highlight advancements in LLM acceleration (e.g., 'Megalodon' for unlimited context), memory optimization ('Headinfer'), and scalable training frameworks ('FlexGen'). She also contributes to federated learning and edge deployment solutions for LLMs.

• Awards:
  • Best Paper Runner-up, ICML 2022
  • Best Paper Award, IISA 2018
  • Best Paper Award, USENIX LISA 2014
  • Rising Star in EECS (MIT 2019, UIUC 2021)

Chen collaborates across academia and industry, focusing on bridging algorithmic innovation with practical hardware constraints. Her lab emphasizes scalable, efficient systems for modern AI challenges.