Haozheng Wan

Haozheng Wan

PHD student at HKU, B.Eng. at HIT

Haozheng Wan （万浩正） Hadrian Wan

Ph.D. Candidate

Department of Electrical and Computer Engineering, CANLab
The University of Hong Kong

I am making AI governable

About me

I am a third-year Ph.D. student in Electrical and Computer Engineering at The University of Hong Kong, advised by Prof. Can Li at the Cognitive and Adaptive Neuromorphic Computing Lab (CANLab). I study efficient, privacy-preserving, and governable AI and agents under real-world computing and security constraints, focusing on hardware-algorithm co-design with analog computing-in-memory (CIM) and device-level stochasticity in emerging memories.

My work uses intrinsic analog errors for hardware security and computational provenance (RGC GRF-funded), regulates them for privacy-preserving computation, and develops monolithic 3D (M3D) architectures (ITF-funded) to overcome analog variability and expand operator support and on-chip bandwidth for fully homomorphic encryption.

Before Hong Kong, I earned a B.S. in Microelectronics Science and Engineering from Harbin Institute of Technology (HIT) in 2023, ranking first in my cohort. I also joined exchange programs at Peking University and the University of Macau, working on AI computing systems and analog IC design.

Research Directions

My research is centered on Making Intelligence Governable: building intelligent systems that are efficient, private, reliable, traceable, and controllable across algorithms, agents, and computing hardware.

AI and agent governance 有跡可循 Studying methods that make intelligent models and agents reliable, traceable, and controllable during generation, reasoning, and tool use.
Privacy-preserving intelligent computation 有密必守 Protecting data, models, and computation through cryptographic and system-level mechanisms for AI workloads.
Hardware-aware resilient AI 有擾必穩 Co-designing algorithms, circuits, and systems so AI workloads remain reliable and secure under hardware variability, faults, and attacks, from training through inference.
Efficient emerging computing substrates 有耗可省 Exploring analog CIM, device-level stochasticity, and monolithic 3D architectures as scalable, energy-efficient substrates for AI computing.

News

  • Candidature confirmed! I am now a Ph.D. candidate.
  • I joined The University of Hong Kong as a Ph.D. student.
  • I obtained my B.Eng. degree in Microelectronics from HIT.
  • I built my own homepage.

Contact

Academic discussion and research questions are welcome by email.