Professor
Department of Electrical and Computer Engineering
National University of Singapore
My research lies at the intersection of computer systems and artificial intelligence, spanning machine learning systems, embodied intelligence, VLSI, and electronic design automation. I have published more than 200 research papers and received several best paper awards and the U.S. NSF CAREER Award.
I received my PhD in Computer Engineering from Princeton University. Before joining NUS, I held faculty positions in three countries: Fudan University in China, the University of Colorado Boulder in the United States, and Queen’s University in Canada. Singapore is my fourth country, and I plan to stay.
Alongside my academic career, I served as Deputy Director and Chief Architect of Intel Labs China and held independent and executive director roles at Concord New Energy Group. I also founded two start-ups whose products have reached customers in more than 40 countries and regions.
Email: li dot shang at gmail dot com
I am recruiting postdocs and PhD students who think creatively and critically, and who want their work to have real-world relevance. If this sounds like you, please email me your CV and a short note on a problem you would like to work on.
“The best way to predict the future is to invent it.” — Alan Kay
My long-term vision is to build personalized embodied intelligent systems — small, continuously learning and self-evolving artificial agents that understand and interact with the physical world. By bridging machine learning algorithms with system software and silicon, I seek to develop adaptive cyber-physical intelligence that co-evolves with individuals over time, serving as intelligent companions that augment human capability, support lifelong growth, and deepen personal understanding of the world.
AI infrastructure is still designed in layers: the model, the software that runs it and the chip are built on separate tracks, so gains that need them to change together are missed. I study how AI agents can design across these layers, from serving software and compilers down to the chip microarchitecture. The agent writes the design problem, solvers search it, and verification and measurement check the result.
Large models keep growing, and so does the cost of training and serving them. I work on making large models run faster and cheaper as they scale, across computation, memory, communication and energy, so that more capable models can be trained and served on the hardware we have.
Intelligence that lives in the physical world has to sense, act and keep learning. I explore small, adaptive agents that learn continually from real-world interaction, connect perception with action, and work alongside people over long periods of time.
Learning to Formulate: Agentic Hardware–Software Co-Design of AI Infrastructure through Progressive Optimization and Verification
Yi Ding, Li Shang. Draft, October 2026.
Not every student will become a great computer scientist, but a great computer scientist can come from any student. I teach every student as if they might be the one.
AI Infrastructure Engineering/Machine Learning Systems
Integration of algorithms, system software, and hardware for building efficient and scalable machine learning systems.
Embodied Intelligence
From perception, world models, and decision-making to vision–language–action models and real robotic systems. This course asks not only what a model can know, but how an agent can observe, reason, and act.
From Rejecting Papers to Discovering Value: Rethinking Scientific Publishing in the Age of AI
The Technology of Discovery: How Intelligence Learns the World
The Trinity of Civilization: Human Capability, Technology, and Institutional Evolution
From Apprenticeship to Algorithms: What Education Becomes When Knowledge Is No Longer Scarce
From Scale to Rhythm: Nietzsche, Underestimation, and the Architecture of Inner Freedom
The Architecture of the Unknown: Philosophy, Wisdom, and the Human Edge
Articulation: Why Knowing Isn’t Understanding
Observe, Don't Touch: Evolution Is the Nature of AI Systems
The Great Academic Re-Sorting: Research, Leverage, and Rare Minds
Prior Discovery: Iterative Falsification in Search of World Structure
From Token Sequences to Structured Knowledge Memory
Agentic System Design Automation for Tensor Computing
Email: li dot shang at gmail dot com