
GitHub
Solo product consultant who delivers full product builds—strategy, design, and code—for early-stage startups. My clients get a fractional CPO, designer, and engineer in one engagement. Previously PM at Mercari and LINE.

Napkin to $4.2K MRR product with paying users in 6 weeks
Early-stage founder had a validated idea for a niche B2B SaaS (invoice reconciliation for SMBs) but no technical co-founder, no design, and a 2-month runway to show traction.
Ship product from napkin to $4.2K MRR with 14 paying customers in 6 weeks
Ran a compressed product process: Claude-assisted user interviews and competitive analysis in week 1, Figma designs in weeks 2-3, and Cursor-built Next.js app in weeks 4-6. Used Claude for spec writing, Cursor for code generation, and Linear for project tracking. Shipped a working MVP with Stripe billing.
Live product with 14 paying customers ($4,200 MRR) by week 8. Founder raised a $500K pre-seed round with the traction.

Checkout redesign recovering $1.2M by cutting cart abandonment
E-commerce client had a 74% cart abandonment rate, significantly above the 69% industry average. Exit surveys pointed to confusion in the 5-step checkout process.
Reduce cart abandonment from 74% to 61% recovering $1.2M revenue
Used Claude to analyze 2,000 session recordings and categorize drop-off reasons. Redesigned the checkout in Figma (5 steps to 2), built a Framer prototype for user testing, then implemented the final version with Cursor. A/B tested against the original for 3 weeks.
Cart abandonment dropped from 74% to 61%, recovering approximately $1.2M in annual revenue. The 2-step checkout also reduced support tickets about order issues by 45%.

Data-backed prioritization framework adopted by 3 startups
Every startup client had the same problem: a backlog of 100+ feature requests and no principled way to decide what to build next. Decisions were based on loudest customer or founder gut feel.
Build feature prioritization framework decreasing cycles 30%
Created a Linear-integrated prioritization system where Claude scores feature requests against weighted criteria (revenue impact, retention risk, effort, strategic alignment). Pulls data from support tickets, churn surveys, and usage analytics. Built a Figma dashboard template clients can customize.
Three clients adopted the framework. Average feature cycle time decreased 30% because teams stopped debating priorities and started executing against a data-backed roadmap.

40-component design system built in 2 weeks with AI-generated code
Client had raised Series A and needed to hire engineers fast, but their codebase had no component library—every page was custom styled. New engineers took 3 weeks to ship their first feature.
Ship 40-component design system in 2 weeks reducing onboarding 3wks→4days
Audited existing UI patterns in Figma, consolidated into a design system with 40 components. Used Cursor to generate React components with Tailwind CSS, Storybook documentation, and unit tests. Integrated with their existing Next.js app.
New engineer onboarding time-to-first-feature dropped from 3 weeks to 4 days. Design-to-code handoff time decreased by 70%.

30-user research study delivered in 5 days vs. 6-week agency quote
Client was pivoting from B2C to B2B but had no understanding of the new target persona. Traditional research agency quoted 6 weeks and $25K.
Complete 30-user research in 5 days vs 6-week agency timeline
Recruited 30 participants through LinkedIn outreach (Claude-drafted messages). Ran 20-minute remote interviews with a Claude-generated discussion guide. Used Claude to transcribe, tag, and synthesize findings into a persona document, jobs-to-be-done framework, and prioritized feature recommendations.
Delivered the full research report in 5 days at a fraction of the agency cost. Three of the four recommended features became the B2B product's core differentiation.

Custom admin dashboard saving 15 hrs/week and cutting errors to <1%
Client's operations team was managing customer accounts through a combination of Airtable, Slack commands, and direct database queries. Manual processes led to frequent errors and 15+ hours/week of admin overhead.
Build admin dashboard saving ops team 15 hours/week with <1% errors
Designed and built a custom admin dashboard in Next.js using Cursor. Claude generated the data models and API routes. Figma for the UI. Connected to their existing Postgres database with role-based access controls.
Ops team recovered 15 hours/week and error rate on account changes dropped from 8% to under 1%. The dashboard became their most-used internal tool.
1 week ago
Added artifact link to 6-week ship play
2 weeks ago
Updated tools list
3 weeks ago