Zongqian Li
Ph.D. Student @ University of Cambridge
Education
University of Cambridge2024.01 – NowPhD in Natural Language Processing (Supervisor: Prof. Nigel Collier)
University of Cambridge2023.10 – 2024.01MPhil in Physics (Supervisor: Prof. Jacqueline Cole)
Queen Mary University of London2018.09 – 2022.06BEng with Hons in Materials Science and Engineering
First Class with Honours, Rank 1/234 in school
Awards
- Google PhD Fellowship Finalist 2026.04
- Qualcomm Innovation Fellowship Finalist 2026.04
- Cambridge International Scholarship, Cambridge Trust Scholar 2024.01
- Distinguished Student Scholarship (Top 0.03%) 2020.12
Work Experience
Google2026.05 – 2026.09Student Researcher
Self-Evolving Coding Agents for Automated Research in Finance, Algorithm Design, and Model Training
ClawAll2026.04 – NowFounder
Multimodal Agent-to-Agent Platform for Shopping, Recruiting, Trading, Healthcare, Legal, and More (X for Agents)
Microsoft Research Asia2025.06 – 2026.04Research Intern (Supervisor: Dr. Shaohan Huang, Dr. Furu Wei)
Data Curation, Reinforcement Learning, and Product Development for Coding Agents
Sequoia Capital China(Hongshan Capital)2025.10 – 2026.04Research & Investment Fellow
Model Evaluation for Financial Tasks, AI Technology Research, Startup Investment
ZhenFund(Top 2 Early-stage VC in China)2023.07 – 2024.01Investment Intern (Mentor: Partner Yuan Liu)
Agents for Startup Sourcing, Strategy Research, Startup Investment, Due Diligence
Deloitte2023.04 – 2023.07NLP Developer (Mentor: Associate Chenglong Li)
ChatGPT Finetuning for Financial Tasks, Digital Consulting Pipeline
Selected Papers
Topics
DataPretrainingFinetuningReinforcement LearningReasoningAgentEvaluationEfficiencySpeech
AutoScientist: Recursive Self-Improvement for Coding Agents in Long-Horizon Tasks Agent
Work in Progress
AutoScientist-Quant: Self-Evolving Coding Agents for Automatic Research in Quantitative Investment Agent
Under Review
Breaking Training Bottlenecks: Effective and Stable Reinforcement Learning for Coding Models Reinforcement Learning
Under Review
Scaling Data Difficulty: Improving Coding Models via Reinforcement Learning on Fresh and Challenging Problems DataReinforcement Learning
Under Review
PT-MoE: An Efficient Finetuning Framework for Integrating Mixture-of-Experts into Prompt Tuning FinetuningEfficiency
NeurIPS 2025
500xCompressor: Generalized Prompt Compression for Large Language Models ReasoningEfficiency
ACL 2025
ReasonGraph: Visualization of Reasoning Methods and Extended Inference Paths ReasoningEfficiency
ACL 2025 (GitHub 500+ Stars)
Flexi-LoRA with Input-Adaptive Ranks: Efficient Finetuning for Speech and Reasoning Tasks FinetuningEfficiencySpeech
ICASSP 2026
Prompt Compression for Large Language Models: A Survey Efficiency
NAACL 2025 (Selected Oral)
Auto-generating Question-answering Datasets with Domain-specific Knowledge for Language Models in Scientific Tasks DataPretraining
Digital Discovery (Q1)