I turn manual operational work into automated systems — 2+ years in consulting operations (Conectys → Accenture), now building AI agents with LangChain and LangGraph. Analysis that becomes automation that saves real hours.
I'm an AI Automation Engineer specialised in building intelligent systems with Large Language Models, AI agents and process automation — backed by 2+ years as a Data Analyst in consulting and multinational operations (Accenture, Conectys).
I design and ship AI and automation systems built to run inside real companies: multi-agent pipelines with LangGraph, RAG document intelligence, LLM-powered workflow automation with n8n, and a self-hosted AI agent that runs 24/7 on my own infrastructure.
My data background means I don't just prototype — I build systems that survive contact with real operations. Legal work permit for Portugal/EU.
Simulated inbox (emails, PDFs, invoices, forms) → LLM-powered structured extraction → validation → Postgres persistence → automatic routing by type and urgency. The LLM is a component of the pipeline, not the product.
View on GitHub ↗Production-ready Retrieval-Augmented Generation system for querying PDF documents in natural language — chunked and embedded with local sentence-transformers, served via FastAPI.
View on GitHub ↗Automated lead processing with n8n + LLMs: classifies incoming leads (industry, size, intent score), enriches them with personalized outreach suggestions and routes by priority — with Slack/Telegram notifications.
View on GitHub ↗Self-hosted AI assistant running 24/7 on my own server: Telegram interface, persistent memory, scheduled automations via systemd/cron (price tracking, notifications) and multi-model routing.
Production-grade Excel VBA toolkit that automates the full daily reporting cycle: consolidates multi-source CSV/Excel exports, cleans and deduplicates data, applies corporate formatting, exports a print-ready PDF and distributes it via Outlook — with centralized error handling, performance tuning and audit logging.
View on GitHub ↗LangGraph multi-agent system: specialised agents parse documents, match them against a criteria profile and score fit (1–10) with structured reasoning — my current deep dive into agentic workflows.
View on GitHub ↗