M.Sc. in Data Science & Engineering · University of Naples Federico II, Italy
Trustworthy AI · Causal Machine Learning · LLM Security Testing
Focusing on bridging Causal Inference, Reinforcement Learning (RL), and LLM security to move red-teaming toward structured, interpretable interventions. Seeking PhD opportunities in safety-critical AI evaluation architectures.
Supervised by Prof. Roberto Pietrantuono (Expected: Sep 2026)
- Approach: Embeds FCI-discovered causal structures into a PPO-based agent for dense reward shaping and action masking.
- Key Results: 61.5% Attack Success Rate (ASR) against Llama-3.3-70B (3.18× improvement over baseline), 41.8% reduced API overhead, and 45.2% ASR on Qwen-3.
- Repos:
causal-rl-breaker&rlbreaker-replication.
- Causal/RL: Causal Discovery (FCI, PAG), PPO, Stable-Baselines3, Hugging Face, Mutator/Judge Architectures
- Systems: Python, PyTorch, Docker, Kubernetes, GPU-aware Containerized Microservices
- AI Systems Engineering Intern @ RESTART Consortium (2025): Deployed GPU-aware Whisper ASR on Kubernetes.
- Software Engineer @ SAAM Webware (2020–2023): Backend microservices.