Luca Pajola, Ph.D., is CTO and AI Engineer at SpritzMatter, a Padova-based AI startup spun out of the University of Padovas SPRITZ security group, where he leads the design and deployment of secure, resilient, and GDPR-compliant AI systems. His work bridges research and production engineering, with a focus on adversarial machine learning, offensive AI, AI red teaming, and Zero-Trust principles for enterprise AI.
Lucas research has exposed real-world vulnerabilities across the AI stack: evasion attacks on hate speech detectors, Unicode-based injections against MLaaS platforms, search engine encoding attacks (Best Paper Award, RAID 2023), and prompt injection risks in RAG data loaders. Several of these findings led to coordinated disclosures with Google, Amazon, IBM, and Microsoft. He is an active red-team pentester for Gray Swan, where he tests agentic AI systems end-to-end and has twice been top-ranked in its LLM jailbreaking competitions (2025).
On the engineering side, Luca has led production AI projects for banking, manufacturing, and document-processing clients, and has coordinated EU Horizon Europe projects: He currently manages SpritzMatters role in SHASAI, a project on secure and trustworthy AI, and previously led REDUCE, a FOODITY open-call project. He has contributed to more than 15 Horizon proposals.
Luca was External Professor at the University of Padova, where he taught Machine Learning for Finance (20242026), and is a regular speaker on AI security, including Securing AI with Zero Trust Principles (Cisco, 2025). He has published at venues such as AISec, EuroS&P, RAID, ICASSP, and ICWSM.
