Our team had the pleasure of actively contributing to the event Secure AI for Health Defense, and Beyond that took place at UC San Diego, on February 19, 2025.

Present our vision for AI-Doctors Without Borders—an AI-powered counterpart to support real doctors in remote and hard-to-access locations. Our approach leverages trustworthy AI agents that enhance collaborative learning while ensuring secure and privacy-preserving AI—eliminating the need to transmit sensitive data.

Moderating a panel on LLM Safety Facets, where experts tackled challenges in scaling LLMs, covering key issues like performance evaluation, fairness, hallucinations, and the risks of AI-generated disinformation and deepfakes.

Demonstrating our Privacy-Preserving Split Learning for LLMs on edge devices, with Alessandro Pegoraro showcasing his research work on a Raspberry Pi using DeepSeek. Participants engaged with the model, experiencing how Split Learning enables high-performance LLMs without the limitations of distilled models—all while keeping data private.

Secure and privacy-preserving AI remains at the core of our research. We continue exploring advanced protection schemes to enable secure collaborative learning without compromising privacy.

A fabulous event:

This event was an incredible platform for discussions on AI safety, security, and healthcare applications, bringing together brilliant minds in the field. A huge thank you to Prof. Farinaz Koushanfar and her team for organizing such an outstanding event!

I also want to especially thank Phillip Reiger, Alessandro Pegoraro from my team, and Nojan Sheybani (from Prof. Koushanfar’s team) for their close collaboration and innovative discussions on collaborative learning, security, and privacy. Looking forward to continuing this impactful collaboration!

For the workshop information see UCSD AI/Safety/Health – Secure AI for Health & Beyond