AI investor · ex-AI engineer · Stanford ICME
I partner with exceptional founders turning AI breakthroughs into enduring companies, bringing technical depth and a long-term perspective.
Currently Director at Hony Capital, previously an AI Algorithm Engineer at ByteDance, and a Stanford ICME graduate.
01
I have always been drawn to the moment when an abstract idea begins to work in the real world.
My journey began at Wuhan University, where studying Financial Engineering first introduced me to AI. I started by applying machine learning to quantitative investing. The encouraging results mattered, but the bigger revelation was that AI's potential extended far beyond predicting markets.
For graduate school, I took a less conventional path and joined Stanford ICME. Its interdisciplinary training across applied mathematics, statistics, and computer science gave me a rigorous understanding of machine learning and shaped the first-principles approach I still use to evaluate technologies today.
After Stanford, I joined ByteDance, one of the world's largest AI-native technology companies, to work on core recommendation systems for Toutiao, Douyin, and TikTok. My work combined Transformer-based sequence modeling, multi-objective learning, and online experimentation to better understand user behavior and long-term engagement at scale.
Stanford's culture of innovation stayed with me. I wanted a broader vantage point on how frontier AI becomes an enduring product and company in the hands of exceptional founders. That conviction led me into investing: to discover, understand, and support entrepreneurs using technology to create meaningful change.
02
My work sits between technical judgment, market insight, and long-term partnership with founders.
2023 - Present
Lead frontier-technology investments across multiple RMB- and USD-denominated funds, with a focus on foundational models, infrastructure, vertical applications, and physical AI.
Led investment execution and portfolio management across 12 companies, working closely with founders from diligence through long-term company building and realization.
2020 - 2022
Worked on core recommendation systems for Toutiao, Douyin, and TikTok, using Transformer-based sequence modeling and multi-objective learning to model user behavior, retention, and long-term engagement.
Built end-to-end recommendation solutions spanning problem formulation, data and feature pipelines, model training, ranking strategy, and online A/B testing, translating model improvements into product impact at scale.
2018 - 2020
Studied at the intersection of applied mathematics and machine learning, with coursework spanning deep learning, reinforcement learning, computer vision, optimization, numerical methods, stochastic processes, and parallel computing.
2014 - 2018
Built a foundation in finance, quantitative methods, and applied modeling, and graduated as an Outstanding Graduate of Wuhan University.
03
Selected investments across multimodal models and applications, physical AI, and semiconductors.
Physical AI · Home Robotics
Building trainable home robots around a data-first loop spanning embodied world models, modular hardware, and user-guided skill learning.
View company →
Multimodal AI · Creative Platform
Kuaishou's AI creative platform, unifying video understanding, generation, and editing into a multimodal workflow for creators and enterprises.
View company →AI Models · Financial Applications
Building agentic investment technology and financial time-series models for the next generation of asset management.
View company →
Physical AI · World Model
Developing general-purpose embodied intelligence systems around robot brains, manipulation intelligence, and scalable deployment.
View company →
Physical AI · Aerial Intelligence
Founded by HKU Professor Fu Zhang, SPARO builds general-purpose aerial intelligence for autonomous drones and physical AI.
View company →
Physical AI · World Model
An embodied-intelligence company developing world models and deployment systems that bring general-purpose AI into the physical world.
View company →
Physical AI · Spatial Intelligence
Developing AI-based 3D reconstruction, spatial perception, and robotics technologies for digital twins and physical-world data.
View company →
Semiconductors · Edge AI
Building edge AI chips and full-stack visual intelligence solutions for consumer electronics, robotics, and smart devices.
View company →
Multimodal AI · Video Generation
Building industrial-grade AI video engines and world models for creators, studios, advertising, entertainment, and simulation.
View company →
Semiconductors · Silicon Photonics
Developing silicon-photonics chips and optical interconnect solutions for data centers, AI computing, and high-speed communications.
View company →
Semiconductors · AI Computing
Developing compute-in-memory intelligent chips and efficient computing platforms for next-generation edge AI workloads.
View company →AI Models · Consumer Applications
A global AI-native technology company building content, creation, and information platforms powered by large-scale recommendation and foundation models.
View company →04
I am interested in how AI-native systems can reshape investment judgment: not only analyze each opportunity more effectively, but carry learning from one decision into the next.
Open source · Investor workflow
This open-source Codex skill is one experiment in that direction: move first-pass judgment earlier, make verification evidence-based and on demand, and turn each BP, transcript, datapack, or Feishu link into a living investment view that improves as new information arrives.
View on GitHubOpen source · Private knowledge layer
This open-source Codex skill adds a private cross-project knowledge layer to AI Work Hub. It links new BPs, notes, datapacks, events, technical themes, and valuation questions to prior project cards, sector maps, valuation anchors, and a running thesis ledger, so each decision begins with accumulated context rather than a blank page.
View on GitHubOpen source · Industry & technology research
Good investment questions often reach beyond a single company. This open-source Codex skill explores the technology, competitive landscape, and market behind an opportunity, working from primary sources, comparing alternative paths, and making assumptions explicit to develop an independent investment view.
View on GitHub05
I enjoy meeting people who think deeply about technology, products, and the future. If you would like to exchange ideas, I would be glad to hear from you.