Ang Li Receives Cisco Research Award

Assistant Professor Ang Li has been awarded a Cisco Research Award for his proposal titled Federated Fine-tuning of Black-box Large Language Models via Proxy Models at the Edge. He will receive a gift fund of $75,000 for unrestricted research purposes.

Cisco Research is a division of Cisco Systems, Inc., a multinational digital communications technology company based in San Jose, California. Its research division partners with academia and research labs through gifts that link new ideas and challenges that are relevant to their core business, with a focus on business, technology and societal impact. Their main areas of focus include AI/ML, computer vision, cybersecurity, natural language processing (NLP), quantum information processing, networking and distributed systems.

Li has noted three challenges to fine-tuning large language models (LLMs), which require specialized data to perform tasks. The first is that privacy concerns and highly dispersed data can restrict the performance of LLMs due to dependency on the quality and quality of domain-specific information.

Next, the time and memory space needed for the substantial computational and memory resources required can be cost prohibitive for many organizations. And finally, the pre-training required for LLMs involves enormous computational resources and high costs, which can lead to restricted access due to intellectual property protections.

To address these issues, Li plans to develop a privacy-preserving system for fine-tuning blackbox LLMs. To do this, he will develop efficient federated LLM fine-tuning algorithms that will address both data privacy and intellectual property protection of LLMs in addition to the resource constraint of the participating devices. Ultimately, this will allow more widespread and secure access to LLM fine-tuning.

Prior to joining ECE in the fall of 2023, Li was a research associate at Qualcomm AI Research. He received a Ph.D. in Electrical and Computer Engineering from Duke University, as well as a Ph.D. in Computer Science from the University of Arkansas. His research interests include federated learning, distributed machine learning, machine learning systems, edge computing, AI for IoT, and trustworthy machine learning.

Published February 18, 2025