Websites and online stores
For businesses and founders. Design, code and Mercado Pago payments, with a clear quote.
For businesses and founders. Design, code and Mercado Pago payments, with a clear quote.
Things that save you hours: reading emails, putting reports together, answering questions. There are demos you can try.
For businesses, founders and agencies who need a site that works.

White-label brewery e-commerce plus ERP-lite. Public storefront and an admin with batch traceability, stock movements, suppliers, purchase orders, POS, cash register, and reports. Mercado Pago Checkout Pro, one deploy per client.

Boutique cabin booking site. Calendar with date-overlap validation, Mercado Pago Checkout Pro, transactional email, AI semantic search.

SaaS membership platform for professionals (yoga, pilates, coaching). Monthly recurring billing with MP Subscriptions, customer portal with real cancel/pause/resume, MRR dashboard.

E-commerce demo for a Patagonian chocolatier. Multi-item cart with HMAC session, Argentine shipping engine (3 zones × 3 carriers), admin panel, transactional email.

Travel agency for Bariloche, Patagonia. 14 real excursions, native i18n (ES / EN / PT-BR), 96 SSG pages, AI-generated FLUX 1.1 Pro hero images, full security audit.
App for Tiendanube stores that decides which shipping and payment options the buyer sees based on the cart (product, category, cart total), with a simulator to test rules. Built under the Mingo brand. In development, pending Tiendanube's partner enablement.
For anyone who wants to see things running before deciding.

RAG evaluation lab. Benchmarks chunking strategies and retrieval configs end to end, with citations grounded by chunk ID and confidence scoring.

Invoice automation. Vision extraction, validation, purchase-order matching, and anomaly detection with Claude Vision plus a deterministic rule engine; human-in-the-loop review queue for anomalies.

Paste a GitHub PR URL. 4 agents review in parallel (security, impact, test gaps, docs). Every finding pinned to file + line with evidence and concrete fix.

Paste a company URL. Verified GTM brief in 30s. Sequential pipeline: scrape → extract → analyze → messaging → verify. Evidence grounding with fact IDs.

AI early-warning system for developers. Scrapes 8 RSS feeds + diffs 5 provider pages daily. Static dashboard with Lighthouse 100/94/96/100. Zero infra cost. 159 tests.
Conversational appointment booking over the WhatsApp Business API with Google Calendar sync. Deployed in production.
B2B sales research pipeline. Company URL → scored outreach draft in 15-30s. Researcher fetches Tavily + homepage scraper + Apollo in parallel. 4 sequential agents.
~$0.08-0.15 / run
Competitive intelligence monitor. Snapshots pages with Playwright, extracts structured data via Claude, diffs against previous snapshots, scores threats. Webhooks fire above threshold.
Open-source CLI. Reads native session logs from Claude Code and Codex CLI, distills them into reusable SKILL.md files. Works with any LLM or none (--no-llm mode).
Joined AirLST, in Munich, as a frontend developer
Stayed at AirLST, remote from Argentina
Started building with AI agents
Finished at AirLST and went out on my own
Frontend developer at AirLST (event technology, Munich), from June 2018 to November 2025. I built each event's site to order: registration form, agenda, hotel, speakers and venue.
1000+ event pages for AirLST clients such as Mercedes-Benz, Netflix and BioNTech.
Four things I do differently.
Every commercial demo runs in simulated mode: you complete a checkout, generate a report, trigger a webhook. Nothing to clone, no keys to set. The AI demos accept your own key if you want the real model.
Every repo has a public BUILD_LOG.md with how long it really took. No "made it in two hours".
I direct the AI to write the code: I decide what gets done and I test it. In the bigger projects, like Cumbre, there are automated tests with Vitest and Playwright.
The projects are on GitHub, with their history. You can see how each thing was put together.
Public repos sorted by last push. Updates every 24h.
Live, GitHub API

I started in 2018 as a frontend developer at AirLST, an event technology company based in Munich. I built each event's site to order: the registration form and the pages with the agenda, the hotel, the speakers and the venue. That came to more than 1000 pages over seven and a half years, with Vue and Nuxt. I lived in Germany until 2021 and then kept working remotely from Argentina.
In 2024 I started using AI to build, and that is how I work today: I direct Claude Code, which writes the code, and I decide what gets done and test it. I finished at AirLST in November 2025 and since then I have been making my own projects, with demos you can open.
I live in Bariloche, in Argentine Patagonia. I have European citizenship (Iceland). I speak Spanish and English, and a little German and Icelandic. If you want us to work together, message me.
What my projects use.