CV

Supervisor: Prof. Pietro Liò (Cambridge)  ·  Co-Supervisor: Prof. Fabrizio Silvestri (Sapienza)

Focus: Graph Neural Networks, Geometric Deep Learning, Topological Deep Learning and Diffusion Models

Memberships: RSTLess Group · CaRAML Group

Teaching Assistant: Geometric Deep Learning, University of Cambridge

Final mark: 110 / 110  🏅 Top marks

Thesis: Polynomial Sheaf Diffusion and Equivariant Sheaf Neural Networks

Memberships: ALCOR Lab · ISPAMM Lab

Courses: AI, Machine Learning, Neural Networks, Robot Programming, Robotics I & II, NeuroEngineering, Computer Vision, Medical Robotics, Reinforcement Learning, Autonomous & Mobile Robotics, Deep Learning, Advanced ML, Generative AI Safety, HRI, Robot Benchmarking & Competitions, Multilingual NLP.

Final mark: 110 / 110 with Honours  🏅 Summa cum laude

Thesis: Building Real-Time Multivariate Anomaly Detection Systems in Industry's 5G Networks — GitHub

Membership: OmniAI Lab

Courses: Deep Learning, AI Lab & Computer Vision, ML, AI, Web & Software Architecture, Cybersecurity, Programming I/II, Databases, HCI, Networking, OS, Algorithms, Statistics, Startup & Management, Computer Architecture, Calculus I/II/III, Linear Algebra, Physics, Probability, Foundations of CS.

Final mark: 100 / 100  🏅 Top marks

Project: Electric Company: Ticketing System — GitHub

Supervisor: Prof. Kazuya Yoshida

Project: MoonBot Navigation — Autonomous lunar robot with Dijkstra-based path planning, visual object detection, custom gripper control.

🏆 Won TESP 2025 Competition 📜 Certificate of Research Excellence

Supervisor: Prof. Alexander Steinmaurer

Project: CareConnect — AI-driven hospital system for environmental data querying, real-time graph generation, and robotic action triggering via LangChain.

🏅 Best Poster & Project Award

• Researched Multivariate Anomaly Detection over 5G time-series.
• Designed a customer-centered & customer-agnostic workflow from network data integration to anomaly identification; validated with a major network customer.
• Reduced models to train/maintain and lowered design/inference costs while preserving metrics.