Hello, I'm
Great Minds Fellow at the University of St. Gallen in Switzerland, working at the intersection of Artificial Intelligence, Software Security, and Technology Innovation.
A computer engineer with a passion for building technology that matters, bridging cutting-edge research with real-world impact.
Luca Di Grazia holds a Great Minds Fellowship ($230,000) at the University of St. Gallen in Switzerland, focusing on the security of AI-generated code, hosted by Prof. Dr. Guido Salvaneschi. Previously, he was a Postdoctoral Researcher at USI Lugano, working on GenAI for software testing and verification under Prof. Dr. Mauro Pezzè.
He earned his Ph.D. in Computer Science (summa cum laude) from the University of Stuttgart, supervised by Prof. Dr. Michael Pradel. His dissertation explored innovative approaches to software maintenance through search and predictions.
During a Research Internship at Uber in Amsterdam, he developed a GenAI-based bug-fixing approach that won an internal competition against 103 teams, with results presented directly to the Uber CEO.
My research interests are on Security of AI Agents (FSE 2026), Generative AI to automatically generate tests (ICSE 2026 and ASE 2025), Python type errors (ASE 2026, ICSE 2024 and FSE 2022, Distinguished Paper Award), and more.
Politecnico di Torino, Italy 
University of Stuttgart, Germany 
Uber, Amsterdam, Netherlands 
USI Lugano, Switzerland 
University of St. Gallen, Switzerland
Secured 189,747 CHF ($230,000) in competitive research funding as Principal Investigator through the Great Minds Fellowship (Universität St.Gallen) for the project Secure Software in the Age of Massively GenAI-Generated Code.
Won Uber's 2023 GenAI Competition in Amsterdam for using Generative AI to improve developer productivity in bug fixing, outperforming 103 teams and presenting the results to Uber's CEO and Executive Leadership Team.
Published at all A* flagship software engineering conferences in the same year (ICSE'26, FSE'26, ASE'26) with three independent research groups; recipient of a ACM SIGSOFT Distinguished Paper Award (FSE'22).
ASE 2026
A large-scale study of Python projects found that third-party libraries are a major source of type errors, and that the full benefits of static typing are only achieved when widely used libraries have near-complete type annotations.
FSE 2026
AgentBound is the first access control framework for MCP servers with a policy enforcement engine that contains malicious behavior without requiring MCP server modifications.
ICSE 2026
E-Test leverages LLMs to identify untested execution scenarios and augments test suites accordingly, achieving an F1-score of 0.54 vs. 0.39 for vanilla LLMs.
ASE 2025
An empirical study using 13,866 unbiased test oracles from 135 Java projects to evaluate LLM effectiveness in generating test oracles.
ACM TOSEM 2025
Envisioning HyperAssistant, an AI tool for 2030 developers. Featured in The Register and Swiss IT Magazine.
ICSE 2024
An automated technique that resolves 85.4% of Python type errors using cross-lingual transfer learning, with high acceptance on GitHub PRs.
ECOOP 2024
A standardized methodology for sampling and studying repositories, emphasizing reproducibility and avoiding unreliable metrics.
ACM COMPUTING SURVEY 2023
A comprehensive overview of 30 years of research on code search: query processing, indexing, retrieval, and ranking techniques.
ESEC/FSE 2022
Analyzing 1.4M type annotation changes from 9,655 Python repositories. ACM SIGSOFT Distinguished Paper Award winner.
IEEE TSE 2022
A search engine retrieving relevant examples from millions of code changes within seconds. 2nd prize at ICSE SRC 2022.
Torstrasse 25
9000 St. Gallen
Switzerland