I'm the engineer you bring in when a running system has to change and can't go down while it does — zero-downtime migrations and rewrites.
Seven years in Rails, plus Go where it has to be fast. I'm usually the person brought in when an old system has to change and can't be down while it does. One step at a time, not a big rewrite weekend.
7+ years, senior
40% faster critical Postgres queries
1,000+ tenants on one platform
~60% less manual bookkeeping
Multi-tenant platform for 1,000+ roofing contractors. I got the critical Postgres queries 40% faster, and built the QuickBooks / Chargebee / Nylas sync so a partner being down just delays things instead of corrupting data. Ran the AWS side too — jobs, CI, deploys.
Rails · PostgreSQL · AWS
Money features for ~10,000 freelancers. FreshBooks and Stripe sync, where a rounding error turns into a support ticket. Built the invoicing and tax automation that cut manual bookkeeping by roughly 60%.
Rails · Stripe · PostgreSQL
Built a CRM from scratch for call centers in Tangier and London — millions of records. Designed the Postgres schema for heavy writes and the APIs behind the live dashboards.
Rails · PostgreSQL
Agritourism marketplace — bookings, Stripe, listings. Made the slow pages fast with better queries and caching.
Rails · Stripe
15+ client sites in PHP, Symfony and Drupal. Where I learned to hit someone else's deadline.
PHP · Symfony · Drupal
App Academy (full-stack web) · OFPPT (computer science)
Ruby on Rails · Go · PostgreSQL · Redis · Elasticsearch
AWS · Docker · Heroku · CircleCI · GitHub Actions
Stripe · Chargebee · QuickBooks
Nylas · Intercom · Postmark · Mixpanel · Slack
Sentry · AppSignal · Rollbar · Skylight
Ember.js · Hotwire · Vue · TypeScript · RSpec · Playwright
How I work
The outages I've dealt with were rarely interesting — usually a missing index, or a webhook that ran twice on retry. That's where I look first.
I'd rather ship something boring that works than something clever I have to babysit. New tools have to earn their place — I've maintained enough of other people's excitement.
Legacy code isn't the enemy. It's running the business, and it usually works for reasons that aren't obvious from reading it once. I read it before I change it, and I change as little as I can get away with.
The goal is a system the next person can understand without me in the room.
Fewer moving parts than the problem seems to need.
The handlers weren't idempotent, so retries applied the same webhook twice. Nobody noticed until the numbers stopped reconciling.
Fix Idempotency keys, replay-safe handlers. Now an outage just delays the sync.
Wrong index, and a query shape that only showed its cost once the tenant filter had real data behind it.
Fix Reworked the queries and the indexes. ~40% off the critical paths. Not clever — just necessary.
Isolation was assumed in the app layer and never actually tested. The kind of bug that's invisible until it's a breach.
Fix Made tenant scoping the default and put it under tests, so it's structural instead of something you remember to do.
Bug reports without making the customer explain. One link records their screen, console and network; support replays it. Rails 8 + Hotwire, a Go service streaming the uploads over WebSockets, a TypeScript recorder and a Chrome extension. Four ways in, one backend.
Rails 8 · Go · Hotwire · WebSockets
A Rails control plane running a fleet of Go agents over gRPC. Per-server routing, Docker builds, deploy logs streaming live. Rails for the product, Go for the load.
Rails · Go · gRPC · Docker
The scary part of a rewrite isn't the code. It's losing customers or data while it happens. Here's how I avoid that.
An old system where one change can break everything — tight coupling, no seams, tests nobody trusts. The reflex is to rewrite it from scratch. That's what kills companies: months frozen, then a new system carrying the old one's bugs.
Cut seams into the running system. Move one slice at a time, behind the live product, with the data intact. Rails stays the product; Go takes the parts that need the speed. Every step ships, and every step can be undone.
The business keeps shipping features while the architecture changes underneath it. No migration weekend, no freeze, nothing the customer sees.
A lot of small, dull, reversible decisions. That's it.
Contact