Huajun Bai

Ph.D. Student @ Tsinghua University

I am a PhD student in the Department of Computer Science and Technology at Tsinghua University, where I am fortunate to be advised by Prof. Jiwu Shu. I am a member of the Storage Research Group.


My research interests lie in Tokenomics, at the intersection of Systems and AI. Currently, I am focused on building efficient inference systems for Large Language Models (LLMs), with a specific interest in speculative decoding.

Huajun Bai

Selected Publications

Specexit: Accelerating large reasoning model via speculative exit Huajun Bai, etc. ICML 2026 Internship @ Tencent [Paper] [Report]
WiP: Efficient Speculative Decoding for AI PCs via Hierarchical N-Gram Retrieval Huajun Bai, etc. MobiCom 2025 (Workshop) [Paper]
PARD: Accelerating LLM Inference with Low-Cost PARallel Draft Model Adaptation Huajun Bai, etc. ICLR 2026 Internship @ AMD [Paper] [Report]
Face: Evaluating natural language generation with fourier analysis of cross-entropy Huajun Bai, etc. NeurIPS 2023 [Paper]
Expressive user embedding from churn and recommendation multi-task learning Huajun Bai, etc. WWW 2023 (Poster) [Paper]
A corpus for reasoning about natural language grounded in photographs Huajun Bai, etc. ACL 2019 [Paper]

Experience

2022 - 2024
Founder & Lead Developer
Early-stage Startup (LLM-based News Processing)

Focused on building autonomous systems for high-throughput news aggregation and summarization using LLMs.

2018 - 2022
Machine Learning Engineer
Recommendation Systems

Developed and optimized large-scale recommendation algorithms and user embedding models.

Education

2024 - Pres.
Tsinghua University
Ph.D. in Computer Science and Technology
2017 - 2018
Cornell Tech
M.Eng. in Computer Science
2013 - 2017
Cornell University
B.S. in Computer Science

Contact

Email: bhj24 [at] mails.tsinghua.edu.cn

Address: F5, Ziqiang Science and Technology Building, Tsinghua University, Beijing, China