Time-sensitive AI systems for physical agents: from DNN perception to LLM planning

The integration of Artificial Intelligence (AI), particularly advanced models like Large Language Models (LLMs), into embodied agents such as robots and autonomous vehicles holds transformative promise. However, a critical yet often-overlooked challenge to realising this vision lies in ensuring that such systems can respond to the physical world in a timely and reliable manner.

In this talk, I will present research on developing time-sensitive AI systems, focusing on Deep Neural Network (DNN)-based perception and LLM-driven planning. I will first introduce a duo-block abstraction that enables timely DNN inference across heterogeneous edge processors. I will then present a time-sensitive LLM serving system using segmented generation and utility-based scheduling to improve responsiveness. I will conclude with future directions in time-sensitive AI systems.

Click here to join the seminar via Zoom
Meeting ID: 984 6038 1155
Passcode: 606340

Dr Neiwen Ling is a Postdoctoral Associate at Yale University, advised by Prof Lin Zhong. She received her PhD from the Chinese University of Hong Kong under the supervision of Prof Guoliang Xing. Her research focuses on time-sensitive AI systems for physical agents, integrating edge computing, machine learning, and real-time systems. Her works have been recognised by top-tier conferences, including a Best Paper Award Finalist (SenSys’22), Best Artifact Award Runner-Up (MobiCom’24), and Best Poster Award (SenSys’22). She is a MobiSys Rising Star and received the N2Women Young Researcher Fellowship. Dr Ling co-founded the FMSys workshop at CPS-IoT Week, serving as TPC Co-Chair(2024) and General Co-Chair(2025). She has served on TPCs for SenSys’25, ICPADS’24, CHASE’23, and the TIoT Distinguished Reviewers Board.

For more information about the ESD Seminar, please email esd_invite@sutd.edu.sg