Engineering systems that last

I’m a student at Sapienza University of Rome enrolled simultaneously in two degree programs: ACSAI (Applied Computer Science and Artificial Intelligence) and Energy Engineering. Two paths that look far apart at first glance, but share the same mental formation: fast problem-solving, goal-directed thinking, clean architecture, reliable solutions, economic efficiency.

The computer science side

I like designing systems and solving problems more than just making things work. I’m drawn to compiled and typed languages, though I use interpreted ones whenever the situation calls for it.

In web development, I’m especially interested in the API design phase — REST and beyond — because that’s where it’s decided whether a system stays organized or becomes a liability six months down the line. Depending on the context, I reach for Go, or Python alternatives like Django (when a full structure is needed) or FastAPI (when something lighter will do).

On the High Performance Computing side, I’ve implemented hybrid MPI+OpenMP and CUDA solutions for computationally intensive simulations, achieving up to 1242× speedup over sequential baselines on GPU and 21× on a 64-core CPU cluster. That work taught me two things: architecture-aware memory management matters as much as algorithm design, and profiling before optimizing is not optional.

I use Linux actively every day, and I’m an unapologetic self-hosting enthusiast. I’m working on making a set of self-hosted services available to the Sapienza student community, built on a 3-node Proxmox cluster running in high availability (HA) with distributed Ceph storage and VPN-based access management.

The AI and data side

I build fine-tuning pipelines for open-weight language models. The flagship project is Manzoni AI: a model that rewrites contemporary Italian into the literary style of Alessandro Manzoni, obtained by fine-tuning Gemma 3 4B via LoRA on a custom parallel corpus I built through distillation from a frontier LLM. The training dataset covers the full text of I Promessi Sposi and was constructed entirely by me, distilling a frontier LLM in the reverse direction and optimizing my use of third-party providers along the way.

On the data analysis side, I’ve worked across Python and R on projects ranging from pure statistical analysis to network science. The most recent is a study of gender bias in movie ratings, combining MovieLens 32M with IMDb cast data to measure — via Cohen’s d — how voter gender systematically interacts with the gender of the film’s lead star. The project included geographic segmentation, occupation breakdowns, and fast-greedy community detection on actor and voter networks.

The energy and physics side

The other half of my path is applied physics: I’m drawn to nuclear solutions, but I’m equally interested in renewables, valuing their distributed and reliable characteristics. My CS skills let me design HPC solutions for engineering and physics problems.

Academic record

I qualified for the national finals of the Italian Physics Olympiad (OLIFIS) in 2022, which granted access to the Physics Summer School in Sigillo. In 2023 I served as team captain for the Italian Mathematics Olympiad (qualifying for the regional Polo training phase, second place as a team), reached the regional finals of the school Chess Olympiad, and stepped in as Institute Representative after the outgoing fifth-year representatives graduated. I graduated Maturità Scientifica in July 2023 with 100/100 cum laude.

I hold the IBM AI Developer Professional Certificate and a full ICDL certification.

In short

I move between two worlds — software systems and energy systems — with the same mindset: understand the problem properly, design before doing, and don’t be afraid to get hands-on with the infrastructure, because learning is doing.