Independent software developer
Product design, engineering, and the infrastructure underneath — built by one person who ships and then stays responsible for it.
// what I do
Web applications from the first sketch to the last migration. Typed end to end, tested where it counts, and shipped on a schedule you can plan around.
Interfaces with a point of view: typeset, spaced, and documented as a system your own team can extend without breaking it.
APIs, data models, queues, deploys. Boring in the best sense — observable, reproducible, and cheap to keep running.
Retrieval, assistants and automation wired into real workflows — with guardrails, evaluation, and a clear answer to what happens when the model is wrong.
// selected work
A representative cut. Full case notes available on request.
fintech / platform
Double-entry ledger and reconciliation for a payments company, replacing nightly batch jobs with an event-sourced core. Close time went from four days to ninety minutes.
health / mobile
Patient records, scheduling and prescriptions for a rural clinic network, designed to stay usable on a two-bar connection.
media / web
Thirty years of reporting, typeset properly, indexed for search, and served as static HTML with no client-side framework.
logistics / ai
Route planning with a natural-language assistant over live fleet data. Every suggestion cites the constraint that produced it, so dispatchers can disagree with the machine.
retail / commerce
Headless commerce for a ceramics studio. Edge-cached catalogue, one queue for checkout, zero downtime at 40× normal traffic.
tooling / dev
Internal tooling: every branch gets a URL, a seeded database, and a teardown timer. Reviewers stopped asking "can you push it somewhere?"
// how it runs
A week of questions. I write down what the thing has to do, what it must never do, and what it costs — in plain language, before any code.
Structure first, then surface. Real copy, real data, clickable before it is buildable. You review it in a browser, not in a deck.
Two-week increments, staging deployed every merge, a short written note every Friday. You always know what exists and what is next.
Docs, tests, and a walkthrough with whoever inherits it. Then either an LTS retainer or a clean cut — your call, no lock-in.
// the toolkit
Nothing here is picked because it is trending. Every entry has a story about the day it saved a project.
// working principles
Scope, cost, timeline. If a date moves, you hear it the day I know — not the week it lands. Estimates are promises with a margin, and the margin is written down.
Backups, migrations, monitoring, the on-call runbook. The parts nobody demos are the parts that decide whether the product survives its second year.
Code a new hire can read, docs that answer "why", and no cleverness that only I understand. The measure of a handover is how quickly it stops needing me.
// next step
Tell me what it has to do and who it is for. You will get a straight answer within one working day — including "this is not a project for me", if that is the honest answer.
Remote-first · UTC+3 · replies in English or French
Ask the assistant
Answers from what is actually on this site.
Hello. I can answer questions about the services, the projects above, the process and the stack — grounded in this page, nothing invented. What would you like to know?