Product strategy · Engineering execution · Shopify growth · AI product · SF Bay Area
I turn ecommerce growth problems into measurable product changes.
I'm a product manager with an engineering background. I have led AI styling products, Shopify conversion work, experimentation, analytics and paid-growth systems—from discovery and specifications through implementation and measurement.
Available for select contract work
Shopify Growth Audit
Funnel, promotion economics, catalog hygiene and attribution, in a written report plus a 20-minute walkthrough. Delivered in five business days.
Five-day delivery
one store · one written report · one walkthrough
Four growth problems, from diagnosis to measurable change
discover · define · prioritize · build · measure · learn
Case 01 · Taelor · Homepage revision
Rebuilding the homepage around four first-visit doubts
The highest-traffic page converted worst. I ranked four doubts a first-time visitor has, gave each its own group of sections and its own click-through metric, and rewrote the copy to say "rental" in one line.
Read the case
Case 02 · Kesha Select · Conversion
Three initiatives that lifted cart-to-checkout 29%
Four in five carts never started checkout. 57K recordings and 15 interviews found a threshold nobody could see, and three fixes that needed no developer sprint lifted cart-to-checkout 29% in two months.
Read the case
Case 03 · Taelor · AI product
Teaching an AI stylist what "good" means
A pairing recommender was rejected almost half the time in its first week. Rules from real outfits, stylist controls and a monthly operating review took approval to 81% and spread use to six or more stylists a month.
Read the case
Case 04 · Kesha Select · Paid growth
Meta ads, planned and measured
Stood up paid social from zero: three clean campaign lines, fixed attribution, a creative playbook and one steering metric. Cost per purchase landed in the top tenth of the food category.
Read the case
Also · Migration study, Kesha Select
Built a pipeline joining four years of community records to Shopify orders. Orders were up 22% after the migration, but only a third of loyal buyers had returned, so the roadmap moved from acquisition to retention, with six sized segments and a new weekly KPI.
Also · Promotion economics
Actual promotion cost was 3.6× what the budget ledger recorded. Replaced a gamified unlock with a capped, non-stacking ladder, approved on a one-page cost case.
Also · Catalog and incident
Thousands of products and over ten thousand variants audited, fixes automated with Shopify Flow. Root-caused a double-discount incident and introduced a coupon naming convention.
Rules I keep
- 01Answer the visitor's first question first.A visitor who does not know what you are will not read why you are good at it.case 01
- 02Write the hypotheses before opening the tool.Evidence should rule things out, not confirm a hunch.case 02
- 03One threshold, not many.Per-campaign minimums create split orders, double shipping and abandoned checkouts.case 02
- 04Define the metric before you argue about it.Written down first, agreed, then measured. Otherwise every review is a debate about the number.case 03
- 05A recommender earns adoption by matching the expert's rules.Encode how the stylists pair before asking them to trust the machine.case 03
- 06One trusted number for paid.Rolling seven-day ad spend divided by store revenue, until attribution is proven clean.case 04
- 07Discounts are the last lever.Non-stacking, capped, with a cost case and break-even before launch.tool below
Log
Tools and prototypes I built
Calculator · below
Promotion break-even
Capped, non-stacking ladder against margin. The model behind the cost case the owner approved.
Prototype · React + Claude Vision
Selfie to looks
Pre-subscription onboarding: selfie upload, closet detection, AI look generation, try-on stub, email-gated gallery.
Prototype · server-side vision
Personal style report
Photo analysis with privacy-safe sharing of the resulting report.
Prototype · React + FastAPI
Voice to Jira ticket
Whisper transcription, GPT-4o drafts from voice, text and screenshots, human confirmation before anything is created.
Promotion break-even · example values, not store data
Break-even
Now · updated Sep 2026
Scoping Q4 at Kesha Select: Google Search and Merchant Center from zero, and a Shopify app-cost audit. At Taelor, moving the AI roadmap from pairing toward a human-reviewed shipment copilot. Open to contract growth and analytics work, and to full-time growth or AI product roles.
Started as an engineer, stayed hands-on
Founding engineer, then product manager at a healthcare startup. Founding PM at Taelor. Now Shopify growth at Kesha Select. I still write the SQL and the spec. Bilingual EN and ZH.
The longer story