Junbao Nan

Junbao Nan

Master of Computational Data Science at Carnegie Mellon University

Education

Master of Computational Data Science, Carnegie Mellon University, 2026 - 2027 (Expected)

Pittsburgh, USA

B.Eng. in Computer Science & Engineering, The Chinese University of Hong Kong, Shenzhen, 2022 - 2026

University of California, Berkeley, Aug. 2024 - Dec. 2024

Publication

Wu P., Li K., Nan J., & Wang F. (2025), Federated In-Context LLM Agent Learning. https://doi.org/10.48550/arXiv.2412.08054

Research Experiences

Berkeley Artificial Intelligence Research Lab, UC Berkeley (Supervisors: Prof. Dawn Song and Dr. Xuandong Zhao)

Undergraduate Research Apprentice, AI Safety Division
Oct 2024 - Dec 2024, Berkeley, USA

  1. Reviewed 40+ papers on LLM reasoning and data synthesis, identifying intersections for improving long-tail knowledge acquisition and reasoning trustworthiness.
  2. Summarized and reported work on LLM post-training and test-time scaling, including CoT, PRM, DPO, MCTS, Self-Critic, Self-Consistency, and Rejection Sampling.

Network and Artificial Intelligence System Lab, CUHK-Shenzhen (Supervisor: Prof. Fangxin Wang)

Research Intern, Federated Learning Group
Dec 2023 - Aug 2024, Shenzhen, China

  1. Designed zero-shot and in-context prompts for GPT-3.5-turbo to select and invoke 10 multimodal tools through LangChain.
  2. Improved prompt design for DeepSeek to strengthen its understanding and summarization of tool scenarios and parameter types.
  3. Designed experiments on federated LLM learning for tool invocation under simulated data non-sharing conditions.

Human-Crypto Society Laboratory, CUHK-Shenzhen (Presently Decentralized Computing Lab, University of Washington) (Supervisor: Prof. Wei Cai)

Research Intern, Decentralized Social Computing (Data Analysis) Group
Feb 2023 - Jul 2023, Shenzhen, China

  1. Implemented a Python web scraper with Selenium and BeautifulSoup to collect and clean CryptoRank data while overcoming anti-scraping mechanisms.
  2. Visualized the collected data with Pyecharts and summarized Web3 research and industry developments.

Internship Experiences

Shanghai Qianshiwei Digital Technology (A High-Frequency Quant Trading Startup)

Quantitative Model Research Intern
May 2026 - Jul 2026, Shanghai, China

  1. Filtered high signal-to-noise intervals from 10 million 500-ms SHFE silver observations over 20 months.
  2. Designed a tokenizer and vocabulary construction method for market data and built an Encoder-Decoder (T5) model for futures price sequence prediction.
  3. Performed representation interpretability analysis and designed a large-scale Agent Loop and Harness for hyperparameter search across 59 hyperparameters.

Binance

Data Scientist (AI/LLM) Intern, AI-Data Services Department
Sept 2025 - Feb 2026, Remote/Singapore

  1. Developed an automated intent-classification pipeline for SFT data, achieving 95% precision, 90% accuracy, and 90% recall.
  2. Produced a million-scale URL-grounded Q&A dataset through case decomposition and embedding-based semantic deduplication.
  3. Built query-clarification data for RAG and an SOP knowledge base for mapping user queries to customer-service procedures.

ByteDance

LLM Algorithm (AI4Code) Intern, Trae-MarsCode Team
May 2025 - Aug 2025, Shenzhen, China

  1. Contributed to Context Engineering for Trae’s code-completion model by merging user interaction data with model completion code data.
  2. Built an automated labeling pipeline with LLM-as-a-Judge, Auto-Prompt, and Majority Vote, improving accuracy to 82% and preparing data for RL/DPO training.
  3. Introduced a critic LLM and knowledge base to iteratively improve judge LLM prompts.

Vitaly Medical Inc.

Machine Learning Intern
Oct 2024 - Feb 2025, Berkeley, USA

  1. Built a fall-detection system using the WEDA-FALL wrist-worn accelerometer dataset.
  2. Trained an ANN model with 98.6% accuracy and converted it to TensorFlow Lite for mobile deployment.
  3. Implemented post-fall inactivity recognition and constructed data to test the full system.

Baidu

Speech Algorithm Intern, Speech Technology Department
Jun 2024 - Aug 2024, Beijing, China

  1. Performed slot normalization and converted intent parameters into function calls for a speech LLM for intelligent cockpit applications.
  2. Used Llama3-8B to generate 45,000 chat samples and cleaned raw data into millions of query-intent pairs.
  3. Developed human-feedback features for a Multi-Turn Task-Oriented Dialogue Model and implemented Streamlit interactions for colleagues.

BNP Paribas (China)

Corporate Banking Intern, CIB Department
Jun 2023 - Aug 2023, Shanghai, China

  1. Collected and analyzed corporate clients’ financial data and ESG reports.
  2. Automated batch searching with Python to support industry research.

Shenzhen NINES Internet Co., Ltd. (Campus Start-up)

Front-end Engineer Intern, Dev Department
Sep 2022 - Dec 2022, Shenzhen, China

  1. Worked with the product team on UI requirements and interaction logic.
  2. Collaborated to develop pages for a WeChat Mini Program using Uni-app and Vue.js.

Project Experiences

Independent AI Products

  1. Rustle: A cross-platform social media search and synthesis platform that turns scattered posts and comments into source-linked, verifiable insights.
  2. MenuX WeChat Mini Program: Enriches multilingual, image-sparse restaurant menus with dish photo previews and culturally grounded explanations using multimodal parsing and image retrieval/generation.
  3. Lobster AI Daily Podcast: An AI news podcast operated entirely autonomously by OpenClaw and distributed on Xiaoyuzhou and Apple Podcasts.
  4. ExamPass: A full-stack AI web app that transforms course materials into print-ready A4 cheat sheet PDFs.

Investment Experience

My personal investing focuses mainly on U.S. technology and AI stocks, as well as cryptocurrencies.

Campus Experience

0xCUHKSZ – The Blockchain Association of CUHK-Shenzhen

Core Builder, Feb 2023 - May 2026

  1. Assisted with recruitment, workshops, and community activities.
  2. Helped newcomers learn about the field and prepare for association activities.

Skills, Languages & Hobbies