╔══════════════════════════════════════════════════════════════════╗
║   $ whoami                                                        ║
║   > Sérgio — Software Engineer, building & breaking things AI     ║
║   $ status --current                                              ║
║   > 🇸🇪 Umeå, Sweden · MSc AI @ Umeå University · open to thesis   ║
║          collaborations with AI-focused companies                 ║
╚══════════════════════════════════════════════════════════════════╝

I'm a Portuguese software engineer with 3+ years building banking and public-sector platforms in Java/Spring Boot and Angular - currently splitting my time between shipping production code and an MSc in Artificial Intelligence in Sweden. I like the boring-but-critical parts of software (the query that silently returns nothing, the flow that fails one step out of ten) almost as much as the AI parts, because both teach you to actually understand a system instead of just using it.

Lately I've been drawn to explainable and trustworthy AI - not just "does the model work," but "can you tell someone why it did that, and should they trust it." That thread runs through most of what's pinned below.

> currently reading logs, writing embeddings, and asking "why" one level too many

Explainable AI that reads a contract and tells you why a clause is risky — not just that it is.

• LLM-based clause analysis with risk scoring and human-readable justifications
• Built to avoid the "black box" trap: every flag is traceable to a reason
• JavaScript · LLMs · Explainable AI

Can a €500 consumer camera replace a €40,000 laser scanner for forest inventory? Got within ~1cm of TLS accuracy.

• Built the full COLMAP-based Structure-from-Motion pipeline: cubemap projection, GPS geo-registration, rig-aware reconstruction
• Systematic experiments across FOV, frame rate, and matching strategy; scale-correction reduced DBH error by ~74%
• SLU (Swedish University of Agricultural Sciences) · Machine Vision course, Umeå University

Generative AI system that puts you in someone else's shoes across 500 years of Swedish history.

• Randomly assigns users a historical identity (e.g. a woman in northern Sweden, 1800s) and generates a grounded narrative about their life
• Built with bias mitigation, source traceability, and explainability as first-class design goals, not an afterthought
• In collaboration with Umeå University & Umeå Municipality's Gender Equality Office

Bachelor's thesis: a greenhouse that manages itself so you don't have to.

• ESP32 sensors → MQTT → AWS IoT Core → Node-RED → MongoDB
• Angular dashboard for real-time environmental monitoring and control
• Turku University of Applied Sciences — published thesis

Multi-armed bandits and a Q-learning agent that actually learns to play Pong.

• Implemented and compared exploration strategies for sequential decision-making
• AI-2, Umeå University

Backend

Frontend

Data & Infra

AI / ML

2025 → now   │  MSc Artificial Intelligence · Umeå University         🇸🇪
2022 → 2024  │  BEng Embedded Electronics & IoT · TUAS                🇫🇮
2020 → 2023  │  BSc Computer Science · ISEC                           🇵🇹  (Erasmus → Finland)

• 🅰️ Angular — Udemy
• 🌐 IoT Fundamentals: Big Data & Analytics — Cisco
• 🔐 IoT Fundamentals: Security — Cisco
• 🔌 IoT Fundamentals: Connecting Things — Cisco
• 🐧 Embedded Linux — Cisco
• 📊 MATLAB & Apps Workshop — ISEC

$ status --looking-for
> AI/software collaborations, thesis partnerships, interesting problems