Origin story

The experiment

Legend has it I am the result of a long-running laboratory experiment conducted by Prof. Domenico Cotroneo and Prof. Roberto Natella, somewhere in the basement of the DESSERT Lab.

For years they applied every fault injection technique they had at their disposal — deadlines, rejections, conference reviews at 3 AM, paper #4 of the month, PhD defenses, missing GPUs, broken pipelines, and the occasional Reviewer 2. A full campaign.

Against all odds, the subject exhibited remarkable resilience: no crashes, no silent failures, and an unusually fast recovery time. The experiment was therefore declared a success and the prototype was deployed in production — first as an RTD-A, and now running in continuous integration at DIETI.

Residual failure mode: a deep, possibly unhealthy, passion for finding ways to break systems — especially AI. It turns out the experiment never really ended; it just moved from the lab to the paper track.

Pietro Liguori with Domenico Cotroneo
With Prof. Domenico Cotroneo — Ph.D. advisor.
Pietro Liguori with Roberto Natella
With Prof. Roberto Natella — co-advisor.
Collaborations

People I work with

Research is never a solo effort. Over the years I have been lucky to build long-standing collaborations with researchers whose work I admire — and, more importantly, whose company I enjoy.

Pietro Liguori with Domenico Cotroneo and Bojan Cukic
Prof. Bojan Cukic — University of North Carolina at Charlotte (USA) A long-running collaboration that started in 2020, when I spent several months in Charlotte as a visiting Ph.D. student during my doctorate. It has since grown together with my advisor Prof. Domenico Cotroneo and produced nine joint publications on AI-based generation of software exploits, shellcode, and robustness evaluation of generative models.
Pietro Liguori with Gabriele Bavota and Cristina Improta at DIETI
Prof. Gabriele Bavota — Università della Svizzera Italiana (USI), Lugano With Prof. Bavota and Cristina Improta (my colleague and friend at DIETI, and frequent co-author) we study Large Language Models for software engineering, focusing on the security and quality of AI-generated code. The photo was taken during Prof. Bavota's visit to DIETI.

Additional active collaborations include Prof. Yulei Sui (UNSW, Australia) on vulnerability detection in LLM-generated code — we recently published together in IEEE TSE — and Prof. Naghmeh Ramezani Ivaki (University of Coimbra, Portugal) on software security and automated detection of insecure AI-generated code.

Project

Redlyne — Security for AI-generated code

Redlyne is a VS Code extension that detects vulnerabilities in AI-generated Python code and proposes one-click patches — directly in the editor, without ever sending code to a server.

It is the natural continuation of my research line on the security of AI-generated code: instead of stopping at the paper, the detection logic ships as a tool developers can actually install and use while they code. Everything runs locally — no code, no telemetry, no metadata leaves the machine.

You can install it from the VS Code Marketplace or browse the source on GitHub.

Startup

ARGO — AI for healthcare

Alongside my academic work, I am co-founder of ARGO, a startup founded together with Nefrocenter to bring AI for healthcare from the research bench to real clinical practice.

ARGO operates at the intersection of machine learning, software reliability, and medical decision support — building tools that help clinicians make better decisions, faster, without compromising on safety, explainability, or trust. We work on real data, with real doctors, on problems that actually matter to patients.

We're hiring

If you are a student, researcher, or engineer interested in AI for healthcare, clinical decision support, or trustworthy ML, we are actively looking for new profiles to join the team — both for research collaborations and for permanent positions.

Possible backgrounds: machine learning, data engineering, medical informatics, software engineering, or any field where you have built something real. If you are curious, get in touch with a short note about what you want to work on.

On the road

innovIT — San Francisco

In January 2026, I was selected by the Italian Innovation and Culture Hub (innovIT) to present DeVAIC — our tool for security assessment of AI-generated code — in San Francisco.

innovIT is the Italian government's outpost in the Bay Area, created to connect Italian research, startups, and industry with the Silicon Valley ecosystem. Being selected meant a week of talks, meetings, and conversations with investors, researchers, and engineers based in the Bay.

This story, I am fairly sure, will have a sequel.

Mentorship

Guiding students

A large part of what I do is not writing papers — it is being available to the young people who are just starting. I supervise theses at Università di Napoli Federico II, I mentor members of the CyberChallenge.IT team as local coordinator, I teach cadets at the Italian Air Force Academy (Accademia Aeronautica) in the Practical Cybersecurity course, and I work closely with the engineers and students joining the ARGO team.

These are very different audiences — undergraduates, master students, CTF players, cadets, junior engineers — but the goal is always the same: help them become independent, ask better questions, break things on purpose, and build something they are proud of.

I also believe strongly in open science: code, datasets, and trained models are released on GitHub (DESSERT Lab) and Hugging Face (OSS-forge). If the work is reproducible, the field gets better faster.