Luis Garbayo FernándezLuis Garbayo FernándezNotes on what I build and what I learn along the way.
Loading game…
The game could not be loaded.
My path has always been guided by the fascination of seeing how technology reshapes the world, a curiosity that today converges into a deep passion for Artificial Intelligence. As I complete my degree in Computer Engineering at the Universidade de Vigo, I have integrated this vision to focus on combining modern web technologies, cloud platforms, and AI. Beyond simply writing code, I focus on the underlying logic and models, always seeking to design scalable, intelligent, and user-centric applications that solve real-world problems.
I have a strong interest in research within the healthcare sector, aiming to streamline existing processes and drive innovative, efficient solutions. This technical and research-driven side is balanced by my background as a STEM educator. Teaching complex subjects to high school students and preparing them for university entrance exams has been vital for my growth. It taught me how to communicate complex ideas clearly, adapt to different learning styles, and maintain a mindset of continuous improvement.
I thrive in dynamic environments where I can blend technical rigor with a collaborative spirit. I don't just seek to make a positive impact; I work to achieve it every day through my projects and continuous learning. I see myself as a proactive problem-solver who values teamwork and constant evolution. My goal is simple: to keep growing as a professional while building digital resources that are not only functional but truly valuable to today's society.
Intelligent corporate document search engine built at HackUDC 2026. Features RAG-powered chat with citations, hybrid semantic + lexical search, OCR document ingestion, and support for multiple AI providers. Built with FastAPI, Celery, Redis, and Qdrant.
RAG chat with citations • Hybrid semantic + lexical search • OCR ingestion • Auto category inference • Multi-provider LLM
AI-powered Galician language tutor built during Hacktoberfest 2025. Features real-time chat with Google Gemini integration, automatic grammar and spelling correction, 5 adaptive proficiency levels (A1–C1), user progress tracking, and a responsive web interface. Built with FastAPI, Reflex, and the Gemini API.
Real-time AI chat • Grammar correction • 5 proficiency levels • Progress tracking • Responsive UI
Multi-agent system that integrates, analyses and renders heterogeneous dental data —CBCT scans in DICOM, STL intraoral scans, PDF reports and images— onto a Gaussian Splatting digital twin of the patient. An orchestrator distributes the work across specialised agents that translate every file into a shared document, and the process is reversible: the system can regenerate STL files and images from the twin itself. Built during the ANFAIA scholarship.
Multi-agent architecture with an orchestrator • Digital twin built on Gaussian Splatting • DICOM, STL and PDF ingestion • MCP server for 3D Slicer • Open source, Apache 2.0
A desktop robot made from a spare Android phone and a PC on the same network: the phone is the body — camera, microphone, screen, motion sensors — and the PC is the brain. Real-time object detection across the 80 COCO classes with YOLOv8n, fully local speech-to-text with Whisper, and semantic reasoning through a Gemma 3 served by Ollama, always grounded in what the camera can actually see. No cloud service ever receives the camera or the microphone.
Phone as body, PC as brain • 80-class object detection • Fully local speech-to-text • Grounded LLM reasoning • 24-hour visual memory
Spatial optimisation algorithm that picks 1,000 optimal locations for elderly care facilities out of Spain's ~36,000 census tracts. It combines social demand —dependent population—, economic viability and spatial constraints to avoid facilities cannibalising each other, and adapts density to how urban or rural each area is. Built during the Akademia programme at Fundación Innovación Bankinter.
1,000 locations out of 36,000 census tracts • Demand, income and saturation in a single index • Haversine distances to avoid overlap • Density adapted to urban and rural areas • Interactive maps with Folium
Autonomous research agent for the terminal that searches, ingests through RAG and summarises scientific literature —3D Gaussian Splatting, the DICOM standard, clinical regulation— into Markdown reports. With Claude's native tool calling, the model itself decides which tools to chain: list the documents at hand, fetch new papers, index them into Qdrant, retrieve what matters and write the report.
ReAct loop with native tool calling • Five tools exposed to the model • RAG over a Qdrant vector store • Paper discovery and indexing • Markdown reports
Multi-agent simulation of an automated warehouse where a fleet of autonomous robots stores and retrieves containers with no central planner. Each robot perceives what comes in, decides locally whether it fits its capacity and races to claim it; mutual exclusion is guaranteed atomically by the environment, so no inter-agent negotiation is needed. When shelves saturate, a deadline triggers the outbound cycle and a transport agent clears the zone.
Claim-based allocation, no central dispatcher • Heterogeneous fleet with different capacities • Critical-zone mutex through a supervisor • Outbound cycle triggered by saturation • Live visualisation with Swing
Java Spring Boot REST API with hexagonal architecture for managing book and author catalogs. Exhaustive validation, centralized error handling, and production-ready decoupled architecture.
Complete CRUD • ISBN validation • Edition management • Hexagonal architecture
Java Spring Boot REST API with hexagonal architecture for advanced project management with AI-powered analysis. Integrates RAG services for task estimation and automatic description generation.
Task and milestone management • AI estimation • Project analysis • Description generation
Angular web application for project visualization and management. Intuitive interface with Gantt charts, milestone analysis, and integrated AI estimations.
Gantt view • Progress analysis • AI estimates • Milestone management
Collaborative task management system for shared projects. Full-stack application that allows teams to create projects, assign tasks, and track progress in real-time.
Project management • Task assignment • Progress tracking • Invitation system
Jul 2026 – Present
Jan 2025 – Present
Sept 2024 – Present
Sept 2025 – Jun 2026
Sept 2025 – Dec 2025
I'm always eager to exchange ideas with other builders, dive into complex technical challenges, or explore emerging trends. If you're looking for a collaborator or simply want to talk code, I'm just a message away.
