Curriculum vitae
CV
Experience
Huawei 2012 Theory Lab
Researcher
Microsoft Research & Ads
Research intern · Information retrieval for ads and language models
Research
Data
Post-training data for foundation models, including supervised fine-tuning, preference optimization, retrieval-augmented data, and synthetic data pipelines. I am also interested in pre-training data scaling laws and data mixture design.
Efficiency
Efficient model serving and deployment, especially MoE inference acceleration, expert-load balancing, token dispatch, DeepEP-style communication, and throughput/latency optimization for DeepSeek and Qwen series.
Retrieval
Large-scale retrieval and recommendation with semantic identifiers, vector quantization, and generative sparse retrieval. Recent work includes million-scale semantic IDs for advertising recall and video-language representations for temporal moment retrieval.
Publications
Publications are being updated. For the latest work or collaboration enquiries, please get in touch.