COMP4901D: Embedded AI Systems (Spring 2026)

Computer Science and Engineering, Hong Kong University of Science and Technology

How to run AI models on mobile or embedded devices? What are the unique challenges in embedded AI systems?

Instructor: Xiaomin Ouyang
Class Time: Wednesday/Friday 3pm - 4:20pm (Room 2304, Lift 17/18)
Email: xmouyang@cse.ust.hk (For course-related queries, please use the subject starting from [COMP4901D])
Office Hour: Friday 4:30pm - 5:30pm, Room 3562, Lift 27/28
Teaching Assistants: Mr. Haoxian Liu (hliueu@connect.ust.hk) and Miss Liyu Zhang (lzhangcx@connect.ust.hk) (For enrolled students: Recorded videos for absent classes can be requested by emailing the TA directly.)

Communication and Announcement: Canvas is the main platform for communication and announcement. For students auditing without official enrollment, please email me your name, student ID, and email address to access the course resources not available here and to stay updated with any course changes.

Course Description: This course will introduce the techniques on deploying AI models on ubiquitous yet resource-constrained platforms such as embedded, mobile, and edge devices. Students will develop a deep understanding of embedded and on-device AI systems, gain hands-on experience in implementation, and learn the practical challenges of applying AI in real-world scenarios.

The major topics include 1) Fundamentals of machine learning and embedded systems; 2) Data and system challenges in embedded AI; 3) AI techniques and their efficient deployment on cutting-edge platforms; 4)Real-world applications, such as mobile agent and smart health. The course structure will primarily consist of instructor lectures, paper presentations and hands-on projects/labs.

Course Objectives: After completion of this course, the students will have an in-depth understanding of challenges in embedded AI systems, and be equipped to deploy AI models, including large language models (LLMs), on embedded, mobile and edge devices for applications such as mobile agents, smart health, and more.

Pre-requisite: (COMP 2012 OR COMP 2012H) AND COMP 2211. Basic understanding of machine learning is required. Familiarity or experience with the fundamentals of embedded or mobile systems are preferred.

Grading Scheme:

• Class attendance & discussion: 10%
• Paper presentation: 10%
• Team Project: 40%
  • Proposal presentation (10%)
  • Midterm presentation (10%)
  • Final presentation (10%)
  • Final report (10%)
• Final Exam: 40%

Reference Materials:

• Conferences: Proceedings of MobiCom/MobiSys/SenSys/UbiComp/IPSN/IoTDI/NeurIPS/ICML/ICLR/CVPR/ACL
• Book: Siam, S.I., et al. (2024). Artificial Intelligence of Things: A Survey. ACM Transactions on Sensor Networks. website

Course Syllabus