I grow products
to millions of users,
then I teach people
how to breathe.

Growth PM. Yoga teacher. 8 years at a consumer platform with 18M+ users and 40M monthly visits. Most of my best work started with the question nobody was asking.

Taipei University of Michigan EN · 中文 · 日本語
Japan Market Lead, DAU 6x
Japan had a large team spending aggressively on paid growth, with little to show for it. I stripped it back to a lean team and organic content.
Six months in, growth was decelerating. I diagnosed it as a direction problem, not an execution problem. More campaign spend wouldn't fix the underlying fit gap. Proposed a retreat: pause paid growth, do deep user research, attend offline events. Found signal in unexpected places: dating and part-time job topics generating spontaneous UGC without any prompting.

When HQ cut paid budget 66%, DAU held steady. The spend had been masking the absence of real pull. I treated the cut as confirmation, not crisis.

Market Entry Team Restructuring Cross-Cultural User Research
User Research Finding
All content on the platform
← 3% was campus-specific
but that 3% drove
41%
of new users'
first interactions
"They weren't looking for trending content. They were looking for their campus."
Web Growth, 40M monthly visits, +30% DAU
The team was improving web as a product. I reframed it as an acquisition channel.
I was the first person to put all acquisition channels on one spreadsheet. That's when I saw it: 40M monthly visits from Google, all staying on web, zero conversion to app. Web users' mental model was "I'm looking something up," not "this is my daily app." So the job wasn't making web as good as app. It was making you want the app.

Tested how much content to show before cutting off. 5 comments was the precise tension: enough discussion to feel real, not enough to feel complete. That number was found with A/B testing, not intuition.

Organic Growth A/B Testing Funnel Design
Before: all comments visible
A
Relationships
I found my best friend's boyfriend on a dating app
❤️ 214 · 💬 83
B1screenshot it RIGHT NOW
B2do NOT get involved, I did this once and lost both of them
B3wait... what if his profile was old?
B4OP check B7's reply, PLOT TWIST
B5this exact thing happened to me last year and...
✓ Read everything, close tab
After: truncated at 5 comments
A
Relationships
I found my best friend's boyfriend on a dating app
❤️ 214 · 💬 83
B1screenshot it RIGHT NOW
B2do NOT get involved, I did this once and lost both of them
B3wait... what if his profile was old?
Read all 83 comments in app →
Content Ranking, +130% new-user posting
The team was creating content for new users. The content already existed. It just wasn't reaching them.
First assumption: not enough content. Tested it, pushed more content, metrics didn't move. Looked closer and realized new users were landing on a generic trending feed full of content that had nothing to do with them. The content they'd care about already existed. The feed just wasn't showing it to them.

New-user posting rate went up 130% after personalized ranking. Not a content problem. A routing problem.

Recommendation ML Collaboration Cold Start Personalization
NTU freshman: default feed
Trending
Celebrity breakup everyone's talking about
❤️ 342 · 💬 128
Lifestyle
Best night markets in Taipei ranked
❤️ 87 · 💬 31
Trending
This week's viral meme compilation
❤️ 512 · 💬 89
No campus content. User leaves.
Same freshman: personalized
NTU
Which GE courses are actually easy A's?
❤️ 156 · 💬 73
NTU
Honest review: dorms vs. off-campus housing
❤️ 89 · 💬 52
Lifestyle
Best night markets in Taipei ranked
❤️ 87 · 💬 31
Reads → comments → comes back next day.
Self-Serve Ads, 0 to 1 in 6 months
SMBs weren't ignoring the ad product. They were structurally excluded by the sales model.
Sales commissions made small deals unprofitable. Built self-serve ads from scratch in 6 months, coordinating eng, data, sales, legal, and PR. The hardest part wasn't building. It was maintaining belief. Midway through, I noticed everyone's anxiety had the same shape: they each had downstream stakeholders asking "what am I supposed to tell my people?" So I split comms into two tracks. Eng and data got "why this decision, where we are." Sales, legal, PR got "next deliverable, when it ships." Same progress, different language for different needs.
0-to-1 Stakeholder Mgmt Revenue Cross-Team

Self-serve gave thousands of small businesses access to an audience the sales model had priced them out of.

Built for the long tail. Turned out better for everyone.
Team Leadership, 13 person org
I thought the manager's job was strategy. It turned out to be mostly people work.
First six months: focused entirely on execution and metrics. Turnover followed. The remaining team was stuck in a negative spiral. Diagnosed: the problem wasn't strategy, it was that nobody understood why we were doing what we were doing.
Started weekly 1-on-1s. Not status updates, genuine dialogue. Explained the why behind every decision. Asked what they wanted to work on, what blocked them, who they worked best with.

The team started self-directing. They proposed a research field trip on their own, not because I told them to, but because they wanted to see what users looked like firsthand. That's when I knew the shift was real.

Team Leadership Org Culture Cross-Cultural Self-Awareness
What I'm Building Now
Yoga AI

Yoga AI Tools

Every yoga class disappears when it ends. Two tools to fix that.

Case study →
Job search

Job Search Command Center

Treated my own job search as a product problem. Multi-agent, cost-conscious.

Case study →
8 years at a consumer platform (18M+ registered users, 40M monthly visits). Grew from first non-engineering hire to Japan market lead. Based in Taipei.

Open to growth PM roles, consulting, and sports tech.

Email LinkedIn Instagram