About Me
Jingjing Li is a Senior Machine Learning Engineer at Roblox, where she develops models for text safety across the Roblox platform. Previously, she was an Applied Scientist at TikTok, where she built multilingual conversational LLM agents for global customer service. Her research interests center on LLM safety, alignment, and reasoning. She received her Ph.D. in Computer Science and Engineering from The Chinese University of Hong Kong.
Education
PhD in Computer Science
The Chinese University of Hong Kong
B.Eng. in Computer Science
Xidian University
Experience
Senior Machine Learning Engineer
Roblox- Working on LLM post-training and evaluation for text safety and content moderation.
- Built lightweight ML models for real-time PII detection, supported by LLM-assisted synthetic data generation and labeling.
Applied Scientist
TikTok E-Commerce- Post-trained multilingual LLMs for global customer service at TikTok Shop, covering response generation, intent classification, and routing.
- Built tool-using AI agents and automated evaluation pipelines for multi-turn workflows, rapid iteration, and deployment.
Applied Scientist Intern
Tencent Lightspeed Studios- Developed LLM-driven autonomous agent system for a game demo, winning 1st place in the company-wide innovation competition.
Applied Scientist Intern
Amazon Alexa AI- Analyzed semantic fidelity in neural data-to-text models. Proposed a novel decoding strategy to improve semantic accuracy and entity correctness (+8%) over state-of-the-art models (BART, T5).
Research Intern
Microsoft Research Asia- Investigated unsupervised text generation in data-scarce settings. Developed an iterative in-place text span editing approach, achieving state-of-the-art performance on unsupervised text simplification (AAAI 2022).
Education
PhD in Computer Science
The Chinese University of Hong KongAdvisors: Prof. Irwin King, Prof. Michael R. LyuB.Eng. in Computer Science
Xidian UniversityNational Scholarship, Ministry of Education of China
Selected Publications
* denotes co-first authorship.
Fast, Slow, and Tool-augmented Thinking for LLMs: A Review.
Frontiers of Computer Science (FCS),
2026.
SeRTS: Self-Rewarding Tree Search for Biomedical Retrieval-Augmented Generation.
Findings of EMNLP,
2024.
CLongEval: A Chinese Benchmark for Evaluating Long-Context Large Language Models.
Findings of EMNLP,
2024.
A Survey of Text Watermarking in the Era of Large Language Models.
ACM Computing Surveys,
2024.
Unsupervised Text Generation by Learning from Search.
NeurIPS,
2020.
Improving Question Generation With to the Point Context.
EMNLP-IJCNLP,
2019.