Six New Projects Are Empowering Early-Career Engineering and Computer Science Faculty

Six teams from the University of California San Diego Jacobs School of Engineering have been awarded funding to accelerate interdisciplinary research collaborations that include an early-career faculty member. The big idea is to empower early-career faculty to build interdisciplinary research collaborations to the point that they are competitive for multi-year external funding. The effort is funded by Irwin Jacobs and his late wife, Joan.

“I am thrilled to announce our newest awardees into this powerful program. I like to call it our ‘two-two-and-two program.’ Two graduate students from two different faculty labs are funded for two quarters in order to get a cross-disciplinary research effort up and running,” said Albert P. Pisano, Dean of the UC San Diego Jacobs School of Engineering and Special Adviser to the Chancellor. “This program allows us to accelerate the research efforts of early-career faculty while also strengthening the collaborative culture of the Jacobs School among both our faculty and our graduate students.”

This year’s cohort of six teams is made up of twelve faculty and twelve graduate students. Each team works together to build cross-discipline research collaborations aimed at generating early results that will make the team competitive for longer-term external funding.

“I am deeply grateful to Irwin and Joan Jacobs for their vision, wisdom and generosity,” said Pisano. “Each year, it is wonderful to see what is possible with dedicated funding for building research bridges across labs at the Jacobs School.”

The 2026 funded collaborations tackle a series of highly relevant challenges: enhancing the security of AI agents; developing more secure software systems; improving diagnosis of bacterial vaginosis in postmenopausal women; developing programmable and printable “living” materials; working toward new approaches for treating congenital heart disease; and increasing the safety of cislunar space operations.

“I am excited to see these interdisciplinary collaborations develop. For each team, I hope this funding opens new doors that ultimately lead to real-world positive impacts,” said Bill Lin, Associate Dean for Research at the UC San Diego Jacobs School of Engineering and a professor in the Department of Electrical and Computer Engineering.

Summaries of the research by CNS faculty members Earlence Fernandes and Deian Stefan are listed below.

Foundations of Sandboxing Computer-Use AI Agents
Prithviraj Ammanabrolu, computer science professor
Earlence Fernandes, computer science professor

AI-powered agents are poised to become a standard way that people interact with computer-based systems such as financial systems, healthcare portals, government services and workplace infrastructure. There is a big challenge, however: the AI models powering these agents remain vulnerable to being tricked by attackers. To address this challenge, computer scientists Prithviraj Ammanabrolu and Earlence Fernandes are leading a collaboration that aims to develop the science and engineering needed to better secure these kinds of AI agents. The team will develop systems-level sandboxing for AI agents with the goal of minimizing the risks from prompt injection attacks. This will involve creating an “AI agent mode” in web browsers, developing dynamic privilege control, and creating security-oriented benchmarking. The project is focused on advancing AI security while also contributing to workforce development at the increasingly important intersection of computer security and AI.

Assertive Trace: A Usable Security Policy Language for Detecting Security Violations Research Plan
Michael Coblenz, computer science and engineering professor
Deian Stefan, computer science and engineering professor

AI agents are finding vulnerabilities at an unprecedented scale in modern software systems, even in systems that are battle-tested and have seasoned security teams. To address this growing challenge, computer science professors Michael Coblenz and Deian Stefan are developing a security approach that aims to ensure that software systems are secure by design — and that they are able to adaptively defend against new waves of AI-assisted attacks. Their strategy builds on the team’s insight that many software attacks, or exploits, can be observed by sequences of events in execution traces produced by systems as they execute. The team will distill this insight into usable tools by developing a security framework called Assertive Trace. The goal is to allow developers to specify high-level security policies on traces as well as enforcement techniques that will ensure their software adaptively defends against new attacks.

Story Written by Daniel Kane – dbkane@ucsd.edu
Full Story: https://today.ucsd.edu/story/six-new-projects-are-empowering-early-career-engineering-and-computer-science-faculty