From the first line to the frontline.
- Seeking
- Design or product engineer. Hands-on IC.
- Experience
- 15 years product and UX, now shipping production code.
- Stack
- Claude Code, React, Next.js, TypeScript, Supabase, MCP.
- Based
- Waterloo, Ontario, Canada · EST
Case Studies
A running record of what shipped. Versioned, dated, and exact - read it the way you would read a changelog, because that is what it is.
- 01Conceived, designed, and built solo. Event-driven analytics that captures the user's voice the moment friction hits.
- 02Claude turns raw user signal into a queryable "what users want and why," exposed to agents over MCP.
- 03Stack: Next.js, TypeScript, Supabase (Postgres + RLS), MCP, Claude, Groq Whisper, Turborepo, Railway.
- 04Cuts dev guesswork around 70% in early tests. Teams stop arguing from opinion and start arguing from evidence.
Teams guess what users want. Chats, logs, surveys scattered across half a dozen tools. Blind spots slow every roadmap call.
Designed the data model and the surface (quotes, trends, action items) in code. Claude clusters feedback and writes the summaries. MCP exposes the whole thing so agents in the loop see the same context the engineers do.
Early teams cut dev guesswork around 70%. Roadmaps sharpen fast. The gap between user signal and shipped change closes.
Agentic AI plus MCP lets one person ship what a five-person team used to need. Building it solo was the proof point for everything that came after.
- 01Inherited a failing crypto loyalty program with 60%+ churn.
- 02Prompted, designed, and built the first AI email engine: dynamic content, brand-matched visuals, generated recommendations.
- 03Architected the AI interaction models and the design system enterprise clients ship on.
- 04Landed Fabletics, validated PMF, drove revenue inside the first cohort.
- 05The pivot became ltv.ai. Co-founded the company and carried the product through to a $5M Series A from Bling Capital.
Crypto hype crashed. Users ignored wallets, loyalty programs died. No personalization, nothing to bring anyone back.
Built end to end. Claude and Opus handled dynamic content generation. The visual flows shipped from code, not a deck. Send orchestration, segmentation, analytics loop all wired by the same person. Prototype in the morning, live users by Friday.
Retention jumped 45%. Fabletics adopted the engine for repeat-buyer flows. Open rates and click-through climbed; revenue followed. The product carried from MVP through a $5M Series A.
AI agents changed what one person can ship. Staying hands-on in code was where the leverage compounded. Still is.
- 01Founding designer-engineer at CSC. Portfolio of e-commerce brands near bankruptcy, no UX function, no testing discipline.
- 02Built the CRO workflow from scratch. Thoughtful A/B tests, stat-sig results, no endless research cycle.
- 03Designed and shipped the prototypes, instrumented the metrics, ran the experiments against revenue lift.
- 04DirectBuy, OneKingsLane, ZGallery and the rest turned around. 30-50% revenue lifts, profitable exits.
Brands bleeding cash. Bad UX, no data loop, no shared way to test anything. Assumptions were killing conversion.
Set up the testing framework. Identified turnaround opportunities, built fast prototypes in code, launched A/Bs tied to revenue and conversion. Design treated as a conversion mechanism, not a finish coat. No interview rounds, no deck theatre.
DirectBuy +35% conversion. OneKingsLane multi-million turnaround. ZGallery +48% AOV. Topline across the portfolio +340% YoY on key metrics. 12.4% conversion against a 2.8% industry average. Stat sig every time.
Speed wins. Data over opinions. The same system that turned those brands around (prototype fast, ship behind a flag, let metrics decide) is the one that runs the solo work today.
Designs
that
ship as code.
Spent fifteen years on the product side: UX, growth, and design across e-commerce and SaaS. When AI coding tools matured, the next step was obvious. Today every prompt, design, and pull request comes from the same person.