AccesibleMad
A conversational AI assistant that answers accessibility questions about the city from open public data, built as a candidacy for the city's open-data reuse awards.
- Problem
- Urban accessibility information in Madrid is scattered across open datasets that ordinary people can't easily query.
- Role
- Sole builder, end to end.
- Stack
- PostGIS, PostgreSQL full-text search, Claude tool-use, Neon, Vercel, plus a security-hardening layer. No vector database, deliberately.
The story
AccesibleMad started from a simple observation, that the information a wheelchair user or a parent with a stroller needs about Madrid is technically public but practically unreachable, scattered across open datasets that nobody outside a data team can query.
I built a conversational assistant that answers accessibility questions from that open public data, grounded with PostGIS and PostgreSQL full-text search and Claude tool-use, deliberately without a vector database so the answers stay traceable to the underlying records, and I deployed it on Neon and Vercel with a security-hardening layer, as a candidacy for the city's open-data reuse awards.
Let's talk
I work where product, technology and business meet, and I like talking to people building real things there. Whether it is to compare notes, a problem worth solving, a project, or a role, my inbox is open.
ContactMore projects
Spacioos
A B2B SaaS that runs experiential-leisure spaces end to end: bookings and experience management, profitability per experience, per-tenant billing with Stripe Connect, calendar and notifications, and multi-location franchise management.
Kouncil, an AI operating system for consulting firms
A central system that coordinates a set of specialist AI agents, each one handling an area of the firm, so a small team stays in sync without the fragile custom code that usually breaks.