Full-Stack Engineer · Next.js · React Native · AI
Building production SaaS from Madrid, Spain

I have spent the last couple of years building SaaS products end to end — database, backend, web, mobile, deployment. Most of them are multi-tenant with payments and real-time features, so a lot of my time goes into Stripe webhooks, migrations, role-based permissions and the kind of stuff that decides whether a product actually holds up in production. Most of them do, with paying customers using them today.

Right now most of my curiosity goes into applied AI, mostly vision-language models and voice agents. Looking for a team where what gets shipped reaches real users and where people actually care about quality.

I got into programming through gaming. I have been coding Java for Minecraft servers since I was fifteen — SurvivalCore (the gameplay core), VelocityCore (the network layer on top of the proxy), VP-API (the internal API shared across every plugin), VP-Sentinel (anti-cheat), VP-DupeGuard (anti-dupe), VP-Staff (moderation tooling). That is where I first learned to build and maintain real software, before this was my job.

• AI-powered WordPress editor — a SaaS where users edit their WordPress sites by chatting with an AI agent. It speaks both Elementor and Gutenberg natively, can create, duplicate and reorganise pages, edit posts/CPTs/WooCommerce/menus, manage SEO meta (Yoast and Rank Math) and the Media Library. The agent runs on the Vercel AI SDK with multi-provider routing across Claude, GPT-5 and Gemini, and exposes around forty tools against a hardened WordPress REST layer. It ships a custom MCP server, an autoparser that learns template patterns from existing pages so the AI can replicate them on new ones, AI image generation with automatic replacement, before/after screenshot diffs for visual verification, a per-page change timeline with grouped undo, and agent evals to catch regressions across Elementor, Gutenberg, FSE and global-style flows.

• Multi-LLM SEO content SaaS — generates SEO-optimised articles by routing across OpenAI, GPT-5, Gemini, DeepSeek and OpenRouter, with DALL-E 3 image generation and one-click publishing to WordPress through Rank Math. Multi-tenant with TOTP 2FA, Stripe credit packs, CSV bulk import for batch generation, role-based access and a PWA front-end.

• 3D artist-discovery platform — an A&R-style platform for emerging artists built around four themed 3D zones with React Three Fiber, Three.js and Drei, animated GLB scenes with Framer Motion, an identity quiz that produces a 12-month strategic plan, and an Express 5 + Better Auth + PostgreSQL/Drizzle backend. pnpm monorepo deployed on Dokploy.

• Local-discovery web app — a personal product I am building so that small businesses near me can get visibility online: a directory in the spirit of Yelp or TripAdvisor but focused on proximity economy. Next.js 16 with App Router, TypeScript, Tailwind v4 + shadcn/ui, PostgreSQL with PostGIS for geosearch, Drizzle ORM, Better Auth with organisations, tRPC v11 and TanStack Query, Leaflet/OpenStreetMap for the interactive maps, and ISR-rendered pages for SEO. Live at localizate.online.

• Time-tracking SaaS with a voice assistant — built around Spain's mandatory workforce time-tracking law (RD-ley 8/2019): clock-in/out with optional geolocation, absences, configurable schedules, role hierarchy (employee, manager, admin, superadmin), correction-request workflows, four-year legal retention and AES-256-GCM encryption of sensitive data. Turborepo monorepo with a NestJS + Prisma + PostgreSQL API, a Next.js 15 + Better Auth web app and an Expo mobile app. A separate FastAPI microservice plugs into Siri Shortcuts and uses OpenAI function calling so employees can clock in or check their hours by voice.

• Multi-tenant SaaS for hospitality — table management, reservations, dynamic floor plans, push notifications and a customer-facing iframe widget. Stripe subscriptions with feature gating per plan, dual auth for staff (web + Expo mobile) and customer accounts, and an admin panel that handles billing, dunning and access blocking when a subscription lapses.

• Custom Odoo modules — production modules for Odoo 18 (with structured migrations from earlier versions) covering specialised inventory and sales flows for mobile-device retail: brand, model and range catalogue, capacity/colour/condition matrices, automatic product naming with duplicate prevention, integration with the standard purchase, sale and stock pipeline, plus its own migration scripts and test suite.

• Real-estate, rental and B2B platforms — listing portals, lead pipelines, a real-estate marketplace, lightweight CRMs, quoting engines, talent hubs, ZPL label generators and customer-portal automations integrated with email and document workflows.

• Computer vision with vision-language models — an anomaly-detection pipeline over video using Qwen2.5-VL on Ollama. Sampled frame-by-frame analysis, severity classification, JSON exports and a Streamlit UI so non-technical operators can run jobs and review results.

Frontend

Backend / Data

Mobile

AI / Python

DevOps & Observability

I am opinionated about a few things: one implementation per feature, errors that surface instead of getting swallowed, and the unglamorous parts — auth, billing, webhooks, migrations — deserving as much care as the visible features. I work in English and Spanish.

Most of what I build is private client work. I want to give more back to the open-source community over time, sharing the things I learn that I think can actually help others.

_{Versión en español}