I build products people use every day. Fintech, healthtech, e-commerce — thirteen years of it, and now with AI.
Lending, payments, care, commerce. The founding roles set the architecture; the rest shipped features people depend on.

A business lending and payroll-advance platform for Philippine SMEs. Set the architecture from day one, then led the move from monolith to microservices as it scaled.

QR PH integration, bringing cashless payments to thousands of merchants across the Philippines.

Care coordination software for home care and NDIS agencies. Shipping scheduling and mobile features, plus serverless microservices and the mobile app.

E-commerce platform for baby and parenting essentials. Led backend infrastructure and re-architected the database for scale, reliability, and security.

A salary-advance platform — employees draw part of the pay they have already earned, released in minutes and repaid on payday. MVP built from nothing, architecture set directly with the CEO.

A CRM for the field sales team — iOS app and backend, built from nothing.

Led a team automating service fulfillment across mobile and web for an enterprise telco client, then took on Head of Data for the web capability team.

A CMS and web app for a national movie ticketing platform, plus a citywide parking app in an earlier stint.

Supported and deployed a bank's Internet Banking System, leading code and implementation reviews.

Predicts your race finish time from a recent result — 5K to marathon, no sign-up.

Ask your Garmin anything. A chat over your own runs, sleep, HRV and training load via MCP — grounded answers, one chart as the proof, no dashboards. Running against my real account; free hosting is next.

An MCP server that gives any MCP client accurate Philippine statutory payroll math — SSS, PhilHealth, Pag-IBIG and BIR withholding — computed from versioned tables instead of guessed from training data.
Two open-source systems I run every working day, and two essays on what that day looks like.

A light software factory for coding agents: PRD in, reviewed draft PRs out. You settle the decisions and approve the plan; agents build unattended, a second vendor reviews, and the line stops at a draft PR. Merging stays human — enforced by hooks, not prompts.

A code review process you install in every repo: Copilot on every push, a light AI review on every PR, and on risky PRs an AI reviewer brief — the change, its risks, where to look first — before two people sign off. AI reads every line; people own the risk.
Thirteen years in, I type less code than I did as a junior and ship more than I ever have. How AI Light Factory and Three-Pass Review fit into one working day: I settle the decisions, agents draft, AI reads every line, and I merge.
Every seat on my agent line gets two settings: which model, and how much it may think. The model sets the price; the thinking is almost free. Why the seats that judge stay on the frontier at high effort, procedure goes to cheaper models or none, and the line still lands near half the naive spend.