Researching Agentic AI Security and Agentic AI for Security with the ETH Agentic Systems Lab at ETH Zurich.
My work focuses on the security of agentic systems as they move beyond text generation and start interacting with tools, memory, external services, execution environments, and other agents.
I did a research internship at Huawei Research / Huawei Data Privacy Lab, working on privacy-preserving authorization and post-quantum anonymous tokens. That work also led to a research paper submitted to PKC.
MSc Cybersecurity Engineering at Politecnico di Torino · GPA 29.48/30 · Graduating October 2026 · Expected 110L/110
Open to research collaborations in agentic AI, AI security, and systems research. If you're working on something interesting in this space, reach out.
I am currently developing research directions around the security of agentic AI systems at the ETH Agentic Systems Lab.
The broader question behind my work is how we can build agents that remain secure and controllable as they gain access to real systems, tools, data, persistent state, and increasingly autonomous workflows.
Co-creator of an open-source Agentic AI career-context layer that gives AI agents grounded, reusable professional context.
VitaeContext provides structured source-of-truth data, reusable agent skills, context routing, CLI tooling, and MCP interfaces so professional context can be reused reliably across agents and workflows.
An architecture-first, source-grounded project built around a simple premise:
you cannot properly secure an agentic system without first understanding the entire system around the model.
It maps how components such as context, memory, tools, identity, execution, observability, evaluation, and multi-agent interaction create new trust and security boundaries.
Agent security is systems security applied to increasingly autonomous software.
Agents interact with operating systems, tools, APIs, memory, credentials, external data, execution environments, and other agents. Securing them therefore requires reasoning about the whole system: trust boundaries, authorization, isolation, execution, observability, provenance, and control.
My background spans systems and Linux security, vulnerability research, authorization and privacy, AI/ML for cybersecurity, and secure software engineering. I now bring those foundations into my work on agentic AI security.
My repositories and projects contain the concrete work behind that background.
- 1st Place — IEEE-HKN International Hackathon 2025
- 57th / 2,000 teams — Reply AI Agents Challenge 2026
- 88th / 2,000 teams — Reply Hack The Code Challenge 2025
- IEEE-Eta Kappa Nu Honor Society
- OWASP GenAI Security Project / Agentic Security Initiative