Skip to content

AI Engineer and Automation Specialist

AI and automation for systems that cannot fail.

Four years building AI, data and process automation for banking, public sector and regulated infrastructure in Madrid.

Matías Jackson Fascioli
4+
years building AI and automation systems
3
public bodies served: DGT, Metro de Madrid, Madrid City Council
~10
real investment portfolios run through BrokerAI
Minutes
to legalise an energy installation at Sanwa, down from months

Where this work has run

  • AccentureAI and full stack
  • DGTDirectorate-General for Traffic
  • Metro de MadridRoute optimisation
  • Ayuntamiento de MadridCloud migration
  • EurocoinPayBank of Spain registered fintech
  • Bitcoin SV AssociationAgents and on-chain infrastructure
  • RE/MAX CollectionSotogrande Consultores
  • Sanwa EnergíaProcess automation

Selected work

Built, shipped, and still running.

Proof of concept for a private banking institution

BrokerAI

An AI assistant that analyses and manages long-term investment fund portfolios: quantitative fund analysis, risk and return assessment with predictive models, and selection and rebalancing algorithms.

  • Applied across roughly ten real portfolios
  • Still advising the institution occasionally
  • Machine learning
  • Quantitative analysis
  • Python
  • React

Bitcoin SV Association

Autonomous agent wallets

An MCP server that lets AI agents hold their own wallets and operate autonomously on-chain, plus open source libraries that automate full blockchain operations for other teams.

  • API-first across multiple MCP servers
  • Every product controllable by an agent
  • MCP
  • Autonomous agents
  • TypeScript
  • Open source

Sanwa Energía

Energy legalisation platform

End-to-end automation of the registration and legalisation of aerothermal and self-consumption energy installations, integrated with the public bodies involved in the process.

  • Processing time cut from months to minutes
  • Backend plus web and mobile integrations
  • Process automation
  • Public-body integration
  • Full stack

DGT, Metro de Madrid and Madrid City Council, at Accenture

Public sector AI

Three years leading AI applied to data analysis for major public sector clients: predictive models, generative AI embedded in client processes, and RAG systems with semantic retrieval.

  • Optimisation models for Metro de Madrid night routes
  • AI for automated recruitment screening
  • Generative AI
  • RAG
  • Optimisation
  • Cloud migration

Own products

Products I built and run myself. The first three are live right now.

  • QRma

    Open

    Own product, in production

    Permanent QR codes for restaurants. The code engraved on the physical piece never changes; the destination behind it is edited in two taps.

    50-operation REST contract, 15 data models · MCP server exposing 14 tools

  • Blok

    Open

    Own product, ready for pilot

    Shared-amenity booking for residential communities. Residents book the court, the gym or the social room from their phone; the manager and the concierge administer it from the same place.

    194 unit and integration tests · End-to-end suite run against a real Postgres

  • TAJO

    Open

    Maintenance operations, deployed

    A light self-hosted maintenance management system covering workshop, warehouse, shift and work orders, with an ETL that migrated the existing spreadsheet inventory into a real database.

    641 inventory rows migrated out of spreadsheets · Barcode scanning and biometric sign-in on the floor

  • Monero

    Own product, in production

    Finance and invoicing for a mixed tax situation: employee and self-employed at once. The business side and the personal side live together without mixing, and logging an expense costs one photo.

    Installable PWA with passkey sign-in · Receipt capture backed by a vision model

Side projects and prototypes

Built and finished on my own servers. They are not commercial products and are not hosted right now, so there is nothing to click through to.

  • Beacon

    Finished

    Meeting-to-action funnel for a small team. Paste the notes, get structured actions, see the dependency graph and who owns what.

    Next.js 15 · Prisma · Claude API · MCP · 84 tests

  • Mike Ross

    Finished

    Retrieval over legal codes, chunked and indexed, with a conversational mode that decides on its own whether a question needs retrieval or a plain answer.

    RAG · tRPC · Prisma · pdf-parse

  • SynthDoc

    Finished

    Research synthesis across a set of uploaded documents, producing one report rather than answers one file at a time. Subscriptions and usage quotas built in.

    Next.js · Document chunking · Cognee · Postgres

  • TestPilot

    Finished

    Turns a set of documents into exams: chunks the source, generates questions from it, and assembles tests with the material behind each answer.

    Next.js · BullMQ · Prisma · Postgres

  • LeadIn

    Finished

    Instagram DM automation feeding a CRM pipeline: keyword triggers on incoming messages, leads through stages, and a consent log so the outreach stays defensible.

    Instagram API · Next.js · Prisma · Consent logging

  • BookBot

    Finished

    White-label conversational booking. A business plugs in its services and availability; customers book over WhatsApp or SMS by talking to it.

    Twilio · BullMQ · Redis · Prisma

  • BizumBSV

    Prototype

    Bizum-style peer-to-peer payments: send and receive in euros with a phone number as the payment ID, settled on-chain underneath.

    BSV SDK · tRPC · Ramp Network · Sentry

  • ModelGen

    Finished

    Generation pipeline that takes reference images and trending formats and produces video from them, tracking each run end to end.

    BullMQ workers · Next.js · Prisma

  • Bitpanda Explorer

    Finished

    A C# and .NET 8 MVC application that integrates the Bitpanda API and visualises the account and market data behind it.

    C# · .NET 8 · MVC

  • Personal trainer bot

    Finished

    Telegram bot that runs an assessment, builds a workout plan from it, logs the sessions and tracks body metrics over time.

    Telegram Bot API · Prisma · Payments

  • BSV Charts

    Finished

    Analytics dashboard over on-chain network data, built in Python and Streamlit for fast iteration on the charts themselves.

    Python · Streamlit

  • Unity dashboard

    Finished

    Small internal dashboard and API with session auth and a separate machine key, so a Unity client could read and write its data without a login.

    Express · Session auth · JSON store

How I build things that other people have to trust.

Regulated clients do not buy demos. These are the standards I hold my own products to, and you can check every one of them.

Tests run against a real database

Not only mocks. Blok ships 194 unit and integration tests plus an end-to-end suite executed twice: once against mocks, once against a real Postgres.

An API contract before the first endpoint

QRma had a 50-operation OpenAPI contract and a 15-model schema agreed before the backend was written. The contract is the spec, and it stays published.

Decisions recorded with their price

Every significant architectural decision is written down with what it costs and what it rules out, so the next engineer inherits the reasoning, not just the code.

Operable once it is live

Monitoring, incident management, contingency planning and recovery. I am accountable for the availability of a fintech registered with the Bank of Spain.

Technical range

AI and machine learning

  • Generative AI
  • LLMs
  • Autonomous agents
  • MCP
  • RAG with pgvector
  • scikit-learn
  • pandas
  • Predictive modelling

Engineering

  • Python
  • TypeScript
  • Java
  • SQL
  • React
  • Next.js
  • Angular
  • Node.js
  • PostgreSQL
  • Elasticsearch

Infrastructure and operations

  • Google Cloud
  • AWS
  • Azure
  • Docker
  • CI/CD
  • Linux
  • Monitoring
  • Incident management
  • DORA
  • MiCA

Who you would be working with

I am Matías Jackson Fascioli, an AI engineer and full stack developer based in Madrid. I spent three years at Accenture leading AI applied to data analysis for the Spanish Directorate-General for Traffic, Metro de Madrid and Madrid City Council, and I now work across two fronts: autonomous agents and on-chain infrastructure at the Bitcoin SV Association, and the critical infrastructure of a fintech registered with the Bank of Spain.

I hold a BSc in Mathematical Engineering applied to Data Analysis. I build and run my own products on my own servers, which is where most of the standards on this page come from.

  • BSc Mathematical Engineering applied to Data Analysis, Universidad Europea de Madrid
  • Spanish, Italian and Uruguayan citizenship. EU work authorisation, no visa required
  • Spanish native. English C2, Certificate of Proficiency
  • Winner of the university programming contest two years running

Questions I get asked

What does Matías Jackson Fascioli do?

He is an AI engineer and automation specialist based in Madrid, Spain. He builds generative AI, machine learning and end-to-end process automation for banking, fintech and public sector clients, with over four years of production experience.

Does he work with banks and regulated companies?

Yes. He built BrokerAI as a proof of concept for a private banking institution and still advises it. He is accountable for the availability of the critical infrastructure of EurocoinPay, a fintech registered with the Bank of Spain, and contributed to adapting that platform to the EU DORA and MiCA regulations.

Can he add AI to a process we already run?

That is the usual engagement. He integrates retrieval over your own documents, autonomous agents and LLM orchestration into processes that already exist, including the API contract, the queue, the audit trail and the handoff back to a person.

What public sector work has he done?

At Accenture he led AI applied to data analysis for the Spanish Directorate-General for Traffic (DGT), Metro de Madrid and Madrid City Council. That included predictive models, mathematical optimisation for Metro de Madrid night route planning, RAG systems with semantic retrieval, and cloud migration.

What is an MCP server, and has he built one?

Model Context Protocol is a standard that exposes systems as tools an AI agent can operate. He has built several, including one for the Bitcoin SV Association that lets agents hold their own wallets and act on-chain autonomously, and one exposing fourteen tools in his own QRma product.

Does he work in English and Spanish?

Both. Spanish is his native language and his English is C2, certified with the Certificate of Proficiency in English. He holds Spanish, Italian and Uruguayan citizenship and has EU work authorisation with no visa required.

Is he available for contract work, and how fast does he reply?

He takes on consulting and build engagements alongside his current roles. Messages sent through the contact form reach him directly and he replies within one working day.

Where is he based and does he work remotely?

He is based in Madrid, Spain, and works both on site in Madrid and remotely across the European Union.

Tell me what is slow, manual, or risky.

If it involves AI, data or a process that should not need a person in the middle, I can tell you quickly whether it is worth building.

I read every message myself and reply within one working day.