About me
What I do
I'm Edoardo Gabrielli. I studied physics at the University of Pisa, focusing on nonlinear dynamics, machine learning and data analysis. There I learned C, FORTRAN, CUDA and Python by writing code to speed up simulations and machine learning computations. Today I work at ETT on software for archives, institutions and museums; I do the digitisation work in collaboration with Scripta, ETT's dedicated line. Besides my work at ETT, I take on independent projects for private clients.
My background
- since November 2025
Software developer, ETT
After Space IT merged into ETT I continue to work on software for archives, institutions and museums: document reading with OCR and vision models, AI-powered knowledge bases, cultural heritage catalogues. Thanks to my focus on product and design I also take on ML engineer and software architect assignments. I learned to build desktop applications with Tauri and Electron.
- March 2024 – October 2025
Backend developer, Space IT
Digital and artificial intelligence solutions for interactive exhibitions and museums. This is where I learned backend development on the job: Django, FastAPI and Vue in real projects, REST APIs (including one serving high-resolution panoramic photos) and small vision models (CNNs) inside web and mobile applications, working with designers, product managers and other developers. I also migrated site statistics from Matomo to Google Analytics 4.
- 2023 – 2024
Researcher, siHealth Photonics
Technologies for image processing and 3D modelling in healthcare. I validated a 3D model of the irradiance of therapeutic lamps, calibrated a spectroradiometer together with the UK Health Security Agency at Harwell Campus and mentored university interns on machine learning projects.
- 2017 – 2024
University of Pisa
Physics studies specialising in nonlinear dynamics, machine learning and data mining.
Tools I use
Python, Django, FastAPI
My everyday language for APIs and services.
Vue, Nuxt, React, Next and TypeScript
Web interfaces, from the public site to the admin panel, with types shared with the APIs.
Docker and Cloudflare
Deployment and self-hosting, tested on a separate environment first.
Postgres
Structured data, even with several sites on the same database.
vLLM and Claude API
Local vision models and assistants built on Anthropic models.
Machine learning
Small vision models trained with LoRA, CNNs and evaluation on unseen data.
Tauri and Electron
Desktop applications: a Tauri (Rust) shell with a Vue interface and an Electron tool for image processing. I learned them while building a digitisation tool.
How I work
Written specifications and decisions, tests, releases tried on a separate environment first, and self-hosting at home.
Secure code
Roles and permissions, personal data kept to a minimum (my analytics store neither the IP address nor the user agent), up-to-date dependencies, automated tests and a trial on a separate environment before every release.