Summary

Ph.D. student in Computer Science at Renmin University of China, focusing on LLM-powered data agents, reinforcement learning, and post-training for open-source LLMs. My research centers on building cost-effective and reliable agentic systems for data-intensive tasks such as data preparation, entity resolution, and autonomous data science. More broadly, I am interested in developing lightweight agents with stronger data awareness to optimize deep research workflows, reduce expensive trial-and-error, and improve the practical efficiency of LLM-based systems.

Education

Renmin University of China, Beijing, China
- Ph.D. in Computer Application Technology (Sept 2023 - Expected June 2028)
- Honors: National Scholarship (The only 2nd-year Ph.D. recipient in the college)
- Advisor: Prof. Ju Fan / Prof. Xiaoyong Du
Chongqing Jiaotong University, Chongqing, China
- B.S. in Computer Science and Technology (Sept 2019 - June 2023)
- GPA: 4.31/5.00
- Honors: National Scholarship (The only recipient in the college), Mingde Scholarship Nomination Award (Top 20 in the university)

Experience

ByteDance, Beijing, China
- Research Intern (LLM Data Agent Systems) (March 2025 - March 2026)
- Led the research and development of DeepPrep, an LLM-powered data agent system for autonomous data preparation.
- Proposed a tree-based agentic reasoning mechanism impacting reliability and steerability.
- Designed a progressive post-training framework combining SFT and multi-turn RL.
- Built large-scale benchmarks and achieved strong performance at up to 15x lower inference cost.
Renmin University of China, Beijing, China
- Research Assistant (Sept 2023 - Present)
- Led research on AutoPrep, a multi-agent data preparation framework.
- Studied cost-effective LLM inference for entity resolution and proposed the BATCHER framework.
- Co-authored research on autonomous data agents, text-to-SQL, and data-centric LLM systems.

Projects

DeepPrep: Data Agent System for Autonomous Data Preparation (March 2025 - March 2026)
- Developed an execution-grounded environment supporting 31 data preparation operators.
- Implemented a full post-training pipeline for open-source LLMs (0.5B to 14B).
- GitHub
AutoPrep: Multi-Agent Data Preparation Framework (2024 - 2025)
- Designed a planner agent with Chain-of-Clauses reasoning.
- Developed programmer agents with tool-augmented code generation.
- GitHub
BATCHER: Cost-Effective LLM Inference for Entity Resolution (2023 - 2024)
- Introduced a batch prompting framework with demonstration selection.
- Achieved 4x-7x cost savings over standard prompting.
- GitHub

Honors and Awards

National Scholarship (Ph.D.), Ministry of Education, China (2025)
National Scholarship (B.S.), Ministry of Education, China (2023)
Mingde Scholarship Nomination Award, Chongqing Jiaotong University (2023)

Skills & Research Interests

Research Interests: Data Agents, Reinforcement Learning for LLM Agents, Post-Training, Cost-Efficient LLM Systems, Autonomous Data Science, Text-to-SQL
Programming & Tools: Python, LaTeX
LLM & Agent Systems: Agentic Reasoning, Multi-Agent Systems, Supervised Fine-Tuning (SFT), Multi-turn RL, Prompt Engineering, Execution-Grounded Feedback
Data Systems: Data Preparation, Entity Resolution, Data Integration, Analytical Workflows
- Supervisors: Prof. Ashish Sharma and A/Prof. Fiona Johnson
Aug. 2023 - Current: Guest Scientist
May 2023 - Jul. 2023: Scientist
- Helmholtz-Centre Potsdam - German Research Centre (GFZ), Potsdam, Brandenburg, Germany
- Duties included:
  - Climate-informed Flood Frequency Analysis under a changing climate
- Supervisor: Prof. Dr. Bruno Merz
Nov. 2015 - Feb. 2018: Research Assistant/Engineer
- Tropical Marine Science Institute, National University of Singapore, Singapore
- Duties included:
  - DSSAT crop modeling of future rice yield in Vietnam under climate change, Singapore-MIT Alliance project.
  - Development of index-based drought insurance for sovereign disaster risk transfer, International Finance Corporation, World Bank project.
  - Impact of climate change on inland and coastal flooding in Singapore, Public Utilities Board (PUB) project.
  - Effectiveness of ABC Waters design features in residential developments, PUB-TMSI-Monash University project.
- Supervisor: Prof. Shie-Yui Liong
Mar. 2015 - Sep. 2015: Intern
- Ingenieurgesellschaft Prof. Dr. Sieker mbH, Berlin, Germany
- Duties included:
  - Development of the Time-Area function model based on QGIS environment for stormwater management.
  - Flood modelling and mitigation of Hafar Al-Batin city in Saudi Arabia.
- Supervisors: Prof. Frank Molkenthin and Prof. Heiko Sieker

Awards & Honors

2021, Faculty of Engineering Postdoctoral Writing Fellowship funded by UNSW
2019, OzEWEX Summer Institute Scholarship
2018, University International Postgraduate Award (UIPA) funded by UNSW
2013, Erasmus Mundus Scholarship Award by European Union (EU)
2012, Outstanding graduate of Hohai University, Nanjing
2010, National Undergraduate Mathematical Contest in Modeling (Provincial Award)

Publications

Few‐Shot Multihop Question Answering over Knowledge Base

Fan, Meihao, Lei Zhang, Siyao Xiao, and Yuru Liang. "Few-shot multi-hop question answering over knowledge base." arXiv preprint arXiv:2112.11909 (2021).

Cost-Effective In-Context Learning for Entity Resolution: A Design Space Exploration

Fan, Meihao, Xiaoyue Han, Ju Fan, Chengliang Chai, Nan Tang, Guoliang Li, and Xiaoyong Du. "Cost-effective in-context learning for entity resolution: A design space exploration." In 2024 IEEE 40th International Conference on Data Engineering (ICDE), pp. 3696-3709. IEEE, 2024.

Talks

Teaching

Languages

Mandarin(Native), English(Fluent), German(Basic), French(Basic)

Skills & Expertise

Statistical hydrology and water resource engineering.
Strong interpersonal skills with a good sense of teamwork.
Programming Skills: R, C/C++, and Python in both Unix and Windows systems.
Rich experience in modeling and GIS, using MIKE, SWMM, DSSAT, and QGIS.

Service & Leadership

Contribute to the First/Second Order Draft of IPCC 6th Assessment Report (FOD-WGII-AR6 and SOD-WGI-AR6) as a group reviewer
AOGS2024 Hydrological Sciences Session main convener: Statistical Methods in Assessing Hydroclimate Extremes Under Changing Climate
International Conference on Hydroinformatics (HIC2024) main convener: SS10 Climate Change Impacts on Urban Flooding: Challenges and Innovative Solutions
Topic Coordinator for a Special Issue of Frontiers in Marine Science
Reviewer Editor:
- Frontiers in Water (Sections: Water and Climate; Water and Hydrocomplexity)
Reviewer for Scholarly Journals:
- Journal of Hydrology, Environmental Modelling & Software, and Weather and Climate Extremes
- International Journal of River Basin Management
- Journal of Applied Meteorology and Climatology
- Frontiers in Water; Water; Sustainability; International Journal of Environmental Research and Public Health

Membership

American Geophysical Union (AGU)
European Geosciences Union (EGU)
Asia Oceania Geosciences Society (AOGS)
International Union of Geodesy and Geophysics (IUGG)
International Commission of Statistical Hydrology (ICSH-IAHS)
Modeling and Simulation Society of Australia and New Zealand (MSSANZ)

Meihao Fan