I'm passionate about building things end-to-end, and I love fine-tuning LLMs, architecting real-time backends, and crafting cross-platform apps. Based in Padua, Italy.
I'm a Software Developer passionate about cutting-edge technologies and solid design practices. I analyze, architect and build robust, scalable solutions, constantly sharpening my analytical and problem-solving skills.
In recent projects I've integrated Artificial Intelligence models based on LLMs, personally handling both initial model development and fine-tuning on project-specific datasets.
Through in-depth work with LangChain, Hugging Face and local inference tooling, I've built end-to-end AI solutions that simplify complex workflows, automate text analysis and improve user experience.
These topics genuinely fascinate me — I keep studying the latest techniques in Natural Language Processing to deliver state-of-the-art results, whether that's a fine-tuned model, a real-time backend, or a self-hosted platform running on my own infrastructure.
Designed and built a budgeting/business-intelligence solution end to end, covering Java back-end development, JS/Angular front-end, and a Telegram Bot integration.
Part of Team Obelix, building Monolith for the Software Engineering course at the University of Padua — enabling interactive "bubbles" inside RocketChat.
Front-desk duties: online form processing, student records, diploma distribution and file organization.
Technical-electronics internship: computer maintenance, hardware technologies, and direct client support.
A selection of self-driven projects — spanning LLM fine-tuning, real-time systems, mobile apps and infrastructure. Most run live on my own self-hosted server.
Fine-tuned Google's functiongemma-270m-it small language model for reliable structured tool/function calling, using Hugging Face TRL and supervised fine-tuning (SFT) on a task-specific dataset. Shipped with a full model card and reproducible training pipeline.
View repositoryOn-premises LLM assistant plugging a React chat UI into a Java ERP system, powered by small 1–1.5B parameter local models served via Ollama/llama.cpp. Hardware-tuned model selection (Qwen 2.5, Llama 3.2) for fast, private Italian summarization under 1.5GB RAM.
View repositoryA playback-first, self-hosted music platform: FastAPI backend, a private resolver microservice, and three native clients — Android (Java/Media3), iOS (SwiftUI/AVFoundation) and web. On-demand HLS streaming with pre-resolved metadata for instant skip/shuffle.
Server-authoritative implementation of Murlan, the traditional Albanian 4-player card game. Zero-trust game state, Redis-backed crash recovery, PostgreSQL match history, and AI bots with three genuinely distinct difficulty strategies.
A tool that downloads a YouTube concert recording and splits it into individually tagged tracks from a timestamped setlist — turning a two-hour live set into a properly organized, per-song album with artist/album metadata attached automatically.
A Telegram bot that watches the National Vulnerability Database in real time and pushes instant alerts the moment a new CVE is published — keeping infrastructure and dependency tracking a step ahead of disclosed threats.
View repositoryA Proxmox-based home datacenter running a full personal cloud, entirely self-hosted: photo library, source control, media streaming, password vault, home automation, network-wide ad blocking and an end-to-end-encrypted notes sync stack — all reverse-proxied with automatic TLS and weekly backups.
This site runs on itI spend much of my free time experimenting in my home lab — designing circuits, soldering boards, and building home-automation systems on Arduino and Raspberry Pi, driven by curiosity about the logic behind smart devices.
To recharge I hike, walk in nature and run outdoors — the energy and focus I need for my projects. At rest I read widely, from non-fiction to fiction, and explore music from jazz and rock to classical.
I believe in the value of exchange and sociality — hosting friends or joining casual chess tournaments lets me test my strategic thinking and share ideas in an informal, stimulating setting.