Junbao Nan

Junbao Nan

I am currently pursuing a Master of Computational Data Science at Carnegie Mellon University, after completing a B.Eng. in Computer Science & Engineering at The Chinese University of Hong Kong, Shenzhen. My interests include LLM post-training, representation interpretability, AI agents, and building production AI systems and products. I enjoy turning research and engineering ideas into tools that solve real user problems.

Current Projects

techweek.wiki — Plan Your SF Tech Week More Easily

San Francisco Tech Week · Oct 5–11, 2026

An in-depth event and people database that helps you decide where to go, who to meet, and how to plan your week.

  • Understand the event and the people: One-line summaries; backgrounds, roles, and profile links across 4,000+ guest and host entries.
  • Find and plan: Natural-language search, availability, registration requirements, and schedule conflicts.
  • Bring your own agent: Connect your AI agent to explore the database and plan for you.

My role: Tech & GTM, with contributions to product and UX.
My Partner: Product & visual design, with contributions to UX and GTM.

Product Experience

techweek.wiki event search and filters
Discover & CompareSearch → shortlist
techweek.wiki event details and speaker backgrounds
Know Who’s TherePeople → informed decisions

System Architecture

techweek.wiki data collection, Events and People tables, universal search, and product architecture

Rustle — Now profitable.

Rustle is a commercial social-research product I am currently bringing to market. It searches public discussions across major social platforms (X, Reddit, TikTok, RedNote, Douyin, etc.) and turns scattered posts, comments, and replies into source-linked, verifiable reports.

Product Experience

Rustle research interface
Research Interface Question → cross-platform search
Rustle delivered research report
Delivered Report Answer → evidence → original sources

Production System Architecture

Rustle production system architecture

Other Projects

  • MenuX WeChat Mini Program: Adds dish photos and cultural context to multilingual restaurant menus. Now serving around 100 users.
  • Lobster AI Daily Podcast: An AI news podcast run autonomously by OpenClaw, from news collection to publishing. Grew to 1000+ followers across Xiaoyuzhou and Apple Podcasts in its first month.
  • ExamPass: Turns course materials into print-ready A4 cheat sheets.

See Projects for more details.

My Internships

Shanghai Qianshiwei Digital Technology: Quantitative Model Research Intern, 2026.5-2026.7, Shanghai, China.

Custom T5 Encoder-Decoder Model for Futures Price PredictionCustom T5 Encoder-Decoder Model for Futures Price Prediction Hyperparameter Search Agent HarnessHyperparameter Search Agent Harness

Binance: Data Scientist (AI/LLM) Intern, 2025.9-2026.2, Remote/Singapore.

Customer Service AI SystemCustomer Service AI System

ByteDance: LLM Algorithm (AI4Code) Intern, 2025.5-2025.8, Shenzhen, China.

Trae CUE Code Completion Model Post-Training Data FlywheelTrae CUE Code Completion Model Post-Training Data Flywheel

Vitaly Medical Inc.: Machine Learning Intern, 2024.10-2025.2, Berkeley, USA.

Baidu: Speech Algorithm Intern, 2024.6-2024.8, Beijing, China.

BNP Paribas (China): Corporate Banking Intern, 2023.6-2023.8, Shanghai, China.

Shenzhen NINES Internet: Front-end Dev Intern, 2022.9-2022.12, Shenzhen, China.

See “Internships” page for more information.

My Research Experience

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

AI Safety Division, 2024.10-2024.12, Berkeley, USA.

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

Federated Learning Group, 2023.12-2024.8, Shenzhen, China.

CUHK-Shenzhen Human-Crypto Society Laboratory (Supervisor: Prof. Wei Cai):

Decentralized Social Computing (Data Analysis) Group, 2023.2-2023.7, Shenzhen, China.

See “Research” page for more information.