Magdalena Onaindia

.net engineer → ai product builder

Kaixo! I'm Magdalena.

Software engineer turned AI product builder.

I ship real products with AI agents — and host them on my own hardware in Bilbao.

See what I've built Get in touch

the portfolio

Things I build

Products, platforms and experiments — most of them built with AI agents, all of them real. Each card shows its build recipe: how much was AI-assisted, and what only a human could do.

  • live

    Self-Hosting Platform

    A PaaS on my own hardware: git push in Bilbao, public URL everywhere.

    • .NET 8
    • gRPC
    • mTLS
    • Docker
    • Traefik
    • Caddy
    AI-assisted 50%Human 50%
    Build recipe

    AI: Claude Code · CLAUDE.md-driven repo

    Human: The tunnel architecture, the mTLS trust model, and every deployment decision.

    Code ↗Read the build story →
  • building

    Restaurant Manager

    Split the bill by QR: everyone pays their share, the waiter watches the table turn green.

    • Next.js 15
    • NestJS
    • TypeScript
    • Prisma
    • PostgreSQL
    • Stripe Connect
    AI-assisted 90%Human 10%
    Build recipe

    AI: Claude Code · phase-gated plan · e2e-first verification · Playwright-checked UI

    Human: The product vision, the split-payment UX calls, and the acceptance bar at every one of the 12 phases.

    Code ↗Read the build story →
  • live

    DeutschMeister

    I'm learning German. So I built my own B2 grammar coach.

    • ASP.NET Core 8
    • React 18
    • TypeScript
    • PostgreSQL
    • pgvector
    • Claude API
    AI-assisted 70%Human 30%
    Build recipe

    AI: Claude Code · Claude API · evals in CI

    Human: The Clean Architecture layering, the grammar curriculum, and the eval gates that decide when AI features are allowed to ship.

    Live ↗Read the build story →
  • building

    TravelApp

    Group trips without the Excel + WhatsApp + Google Docs chaos.

    • Next.js
    • TypeScript
    • Drizzle
    • PostgreSQL
    • Docker
    AI-assisted 80%Human 20%
    Build recipe

    AI: Claude Code · spec-driven workflow · AI audit reports

    Human: The product spec, the permission model, and saying no to every feature outside the MVP.

    Live ↗Read the build story →

the workflow

How I build

  1. Spec it

    I write the spec and the plan. The product thinking is the part that can’t be delegated.

  2. Build it with agents

    AI agents implement batch by batch. I review every line that crosses an architecture boundary.

  3. Own it

    Everything deploys to my own self-hosted platform — a PaaS I built on my own hardware.

  4. Ship it

    Public URL, real users (often just me — the most demanding one), and iteration.

The interesting question isn't “did AI write it?” — it's “who decided what to build, and does it run?”

the story

About

I grew up in Bilbao and studied electronic engineering at the University of the Basque Country — my thesis taught an NVIDIA JetBot to harvest crops with computer vision. Then I went deep into .NET: document protection at a cybersecurity company in Bilbao, then critical infrastructure for the public sector — health, justice and emergency services.

Somewhere between a master's degree and a home server, the job title stopped fitting. Now I build products end to end: I write the spec, direct AI agents through the implementation, and ship to a self-hosting platform I built myself. Engineer by training, product builder by compulsion.

  • Languages Basque & Spanish (native) · English (C1) · German (B1 in progress — see DeutschMeister)
  • Day job .NET software engineer — Blazor, ASP.NET Core, the works
  • After hours AI agents, my tower, and whatever the next spec says

say hi

Contact

Building something with AI agents? Hiring product-minded engineers? Let's talk.