I am a Distinguished Researcher at Huawei Technologies Canada, where I lead research and engineering on cloud-native AI infrastructure for large-scale post-training and inference of foundation models. Before joining Huawei, I was a postdoctoral research fellow at Stanford University. I received my Ph.D. in Operations Research from Simon Fraser University.
My current research focuses on the following directions:
Building cost-efficient, cloud-native systems for post-training and inference of large language and multimodal foundation models, including supervised fine-tuning, reinforcement learning, and distributed serving. My team contributes to open-source infrastructure such as AReaL, vLLM-Ascend and RLinf.
Applying foundation models to challenging real-world problems in operations research, analytical databases, and large-scale data ecosystems. This direction is reflected in recent work on optimization modeling, text-to-SQL, query federation, and efficient serving of large multimedia models.
Advancing trustworthy and responsible AI through model and data watermarking, federated learning, privacy-preserving learning, and secure data/model management. Recent publications on language-model watermarking, dataset watermarking, fair federated learning, and related topics are representative of this line of work.
For a broader overview of my work, please visit the Publications page. I am always open to discussions on research collaboration, student supervision, and industry-academia partnerships. Feel free to contact me via email at michaelzhang917 at gmail dot com.