I'm studying Artificial Intelligence & Data Analytics at the University of Trieste. I'm interested in computational complexity, formal methods, and systems programming. I like implementing things to understand how they work, then finding ways to check that they do what I expect.
-
Tiling Foundry — implements the Yang–Zhang reduction from CM1-in-3 SAT to a fixed 23-tile Wang system, with a C reference solver, Z3 checks, witness verification, and reproducible run dossiers.
-
TensorForth — a C17 interpreter for a Forth-like tensor language, with OpenMP, reference-counted tensor storage, tests, and Valgrind checks.
-
ADS — a C library where I implement algorithms and data structures by hand to learn them, from binary search trees to graphs and stacks for depth-first traversal. It grows one piece at a time.
-
Cantina DB — a relational database and management app for winery inventory, with trigger-maintained stock, stored procedures, role-based access, and a Streamlit demo. Try the demo.
-
FlowScore — explores credit scoring for gig workers. I designed the architecture, synthetic-data engine, and shock scenarios to compare static credit scores with debt outcomes over time.
-
KaleidoSpeech — a team proof of concept built for a hackathon in Hangzhou, exploring pronunciation training through browser audio, phoneme alignment, and visual feedback.
PAP Render is a pixel-art renderer in Python and NumPy.
I've also worked on team hackathon projects: insurance coverage recommendations with Generali, demand forecasting and supply-chain optimization with Zucchetti, and e-bike energy optimization at EESTech.
Gelsomino Vine is an early experiment toward a local-agent system. I'm working on the protocol layer: deterministic messages, causal ordering, and conformance tests.