Early-stage AI investor at Llama Ventures. Llama Ventures 早期 AI 投资人。
I back founders building the AI that quietly rewrites how people think, learn, and make decisions. Berkeley math, Harvard learning design — I look for products where the underlying technology and the humans using it both get sharper at the same time. 我支持那些正在重新书写人们如何思考、学习与决策方式的 AI 创业者。Berkeley 应用数学，Harvard 学习设计——我寻找的产品，是那种用得越久、底层技术和使用它的人都同时变得更敏锐的产品。
About关于
I’m an investor at Llama Ventures, a $300M+ early-stage fund focused on AI. My job is straightforward: find the founders building things that actually matter, and back them before the rest of the market wakes up.
I came in sideways. I studied Applied Mathematics at UC Berkeley, then spent a year at Harvard studying how humans actually learn — not how textbooks pretend they do. In between I’ve done strategic investing at OPPO, research at Harvard Business School and the Kennedy School, and co-founded a company of my own at Eagon.
That mix is the whole point. Math taught me to take systems apart. Learning design taught me that the most powerful technologies are the ones that make people more capable, not more dependent. I’m drawn to AI companies built on that same instinct.
我是 Llama Ventures 的投资人。我们是一支 3 亿美金以上、专注 AI 的早期基金。我的工作很简单：在市场反应过来之前，找到那些真正在做有意义事情的创业者，并支持他们。
我是从侧面进入这一行的。在 UC Berkeley 学 应用数学，然后在 Harvard 用一年时间研究"人到底是怎么学习的"——不是教科书写的那种。中间在 OPPO 做过战略投资，在 Harvard 商学院和肯尼迪学院做过研究，自己也在 Eagon 联合创办了一家公司。
这种组合就是核心。数学教会我如何拆解系统。学习设计让我相信：最强大的技术，不是让人更依赖，而是让人更有能力。我被那些有这种本能的 AI 公司吸引。
Experience经历
From Berkeley problem sets to Harvard research labs to an investment desk in Palo Alto — every stop has been about the same thing: figuring out how technology and people actually fit together. 从 Berkeley 的习题课，到 Harvard 的研究实验室，再到 Palo Alto 的投资台——每一站都在做同一件事：搞清楚技术和人到底怎么结合。
Llama Ventures · Full-timeLlama Ventures · 全职
Early-stage AI deals across infrastructure, vertical agents, and applied AI. Sourcing, diligence, and partnering with founders from pre-seed through Series A. 早期 AI 项目——覆盖基础设施、垂直 agent、应用 AI。从 sourcing 到 diligence，从种子前一路陪伴到 A 轮。
Eagon LLC
Building from the operator side. The fastest education in what founders actually need from their investors. 在创业者那一侧亲自动手。这是我学得最快的一课——"创业者真正需要投资人提供什么"。
Harvard Kennedy School Executive Education Harvard 肯尼迪学院 · 高管教育项目
Quantitative research supporting executive programs — the kind of work that taught me how to turn messy social data into something a decision-maker can actually use. 为高管项目做定量研究——这段经历教会我，如何把混乱的社会数据，变成决策者真正能用的东西。
OPPO · Full-timeOPPO · 全职
First exposure to writing checks. Strategic investments across hardware, AI, and consumer tech from OPPO’s Palo Alto innovation arm. 第一次真正写支票。在 OPPO 的 Palo Alto 创新中心，覆盖硬件、AI、消费科技的战略投资。
Harvard Business School · Part-time Harvard 商学院 · 兼职
Case research and quantitative work alongside faculty studying technology, strategy, and emerging markets. 和研究科技、战略、新兴市场的教授一起做 case 研究和定量工作。
Education教育
M.Ed. — Learning Design, Innovation & Technology · Graduate School of Education 教育学硕士 · 学习设计、创新与技术 · 教育研究生院
2021 — 2022
B.A. — Applied Mathematics & Data Science 学士 · 应用数学与数据科学
2017 — 2021
Thesis投资理念
Three lenses I use when I meet a founder. They’re overlapping more than mutually exclusive — the best companies tend to be sitting on at least two. 见 founder 时我会用三个视角。它们彼此重叠多于互斥——最好的公司，通常至少占两个。
01 / Reasoning01 / 推理
The boring framing is “copilots.” The real framing is: which tools make their users measurably better at their craft over twelve months of use? That’s the moat. That’s the wedge against incumbent workflows. 无聊的说法是"copilot"。真正的问题是：哪些工具能让用户在 12 个月内，可衡量地变得更擅长自己的专业？那才是护城河，那才是撬动既有工作流的杠杆。
02 / Vertical agents02 / 垂直 agent
I’m more interested in agents that finish work than agents that demo well. The unsexy verticals — back-office, regulated industries, ops — are where I think the next category-defining outcomes live. 比起 demo 漂亮的 agent，我更感兴趣那些真正把工作做完的 agent。那些不性感的赛道——后台、强监管行业、运营——才是下一个定义品类的地方。
03 / Infrastructure03 / 基础设施
Memory, evaluation, orchestration, deployment, trust. Anyone shipping agents at scale today is taping it together. Whoever builds the canonical primitive for one of those layers owns a generation of buyers. 记忆、评估、编排、部署、信任。今天所有规模化跑 agent 的人，都是拿胶带粘起来的。谁能在某一层做出标准原语，谁就拥有了一整代买家。
Selected精选
Recently joined Llama Ventures, and the deals I’ve led are still in the quiet phase. This space is where they’ll show up — founders welcome to ask about specifics. 刚加入 Llama Ventures 不久，我主导的项目还在静默期。这里会陆续放出——欢迎 founder 私下问具体细节。
Five months into the seat at Llama Ventures. A handful of bets placed, more in motion. I’ll publish them here as the founders are ready — until then, happy to talk specifics privately. 在 Llama Ventures 已经 5 个月了。已经下了几注，还有更多在路上。等 founder 准备好了我会陆续公开——在那之前，欢迎私下聊具体细节。
Writing文章
I think best by writing things down. Most of what I publish lives on LinkedIn for now — longer pieces will live here. 我最好的思考都是写出来的。目前大部分内容发在 LinkedIn 上——更长的文章会放在这里。
Now当下
A snapshot of what’s actually on my desk this month — inspired by Derek Sivers’ nownownow. 这个月真实摆在我桌上的事——灵感来自 Derek Sivers 的 nownownow。
Looking hard at vertical AI agents in regulated and back-office verticals — the spaces where the demo-to-revenue path is shortest and the moats are real. 重点在看强监管和后台场景里的 垂直 AI agent——这些是 demo 到收入路径最短、护城河最真实的地方。
Co-running Eagon on the side. The operator perspective keeps me honest when I’m sitting across the table from a founder. 同时还在做 Eagon。这个 operator 视角让我跟 founder 对面坐下时，能保持诚实。
Recently moderated the Builders of Tomorrow AI Super Hackathon and spoke on the VC panel at Founders & Funders SF. Genuinely the best part of the job is finding the next person worth backing in a room full of them. 最近主持了 Builders of Tomorrow AI Super Hackathon，并在 Founders & Funders SF 的 VC panel 发言。坦白说，这份工作最棒的部分，就是在一屋子值得投的人里，找到下一个。
Currently working through the post-AGI economics literature and an embarrassing number of pre-seed decks. Inbox is open. 正在啃 post-AGI 经济学的文献，以及多到不好意思的 pre-seed deck。Inbox 永远开放。
Contact联系
Cold inbound from founders is genuinely my favorite part of the week. Pitch in three sentences — I’ll read it. Founder 直接发来的冷邮件，是我一周里最喜欢的部分。三句话讲清楚你在做什么——我会认真看。