The focus of the UC San Diego Center for Machine-Intelligence, Computing and Security(MICS)  is pioneering the next era of AI revolution by deeply integrating hardware, software, and data. Moving beyond traditional machine learning, our work focuses on the full lifecycle of modern intelligent systems—from the development and efficient deployment of massive generative foundation models and large language models (LLMs), to iteratively leveraging AI (which itself is fueled by the hardware capabilities itself) to dramatically accelerate chip design, to the rigorous validation of AI outputs through proofs of provenance and computation. We are fundamentally re-architecting computing stacks to support these data-intensive models, and using data-intensive models to make better compute systems, ensuring that training and inference can scale intelligently from the cloud to the most resource-constrained edge devices.

Our innovations directly enable the new wave of physically embedded intelligence, where intelligent agentic-based decision-making must occur in the real world with physical constraints and just in-time within physical embodiments such as drones, medical devices, and autonomous systems. We lead in agentic, automated performance optimization that adapts across diverse platforms and have pioneered the safe training and execution of deep learning-based networks directly on edge devices. Furthermore, our holistic approach to cyber-physical systems integrates advances in privacy-preserving computation and verifiable AI to secure these agents against adversarial threats, ensuring that the deployment of generative models in the physical world is both robust and trustworthy.