Agentic Systems for Software Engineering and Security Applications
Wenbo Guo
Computer Science
UC Santa Barbara
Time: 11:15 am - 12:30 pm PT
Date: 10 December 2025
Location: Harold Frank Hall, room 1132, UC Santa Barbara campus
YouTube: https://www.youtube.com/watch?v=tvuHSNu0ngM
Abstract:
In this talk, we will present our recent work on developing agent systems for software engineering and security applications, as well as our training recipes for training small reasoning models under our agentic systems. In addition, I will introduce novel RL-based training methods for improving LLM in coding-related tasks. Furthermore, I will discuss our work in agent security and trustworthiness, including black-box red-teaming and system-level blue-teaming. Finally, I will conclude the talk with recent enterprise and startup experience.
Wenbo Guo is an assistant professor and Zhu Chair at the UCSB CS department. His recent research centers around AI agents and agentic LLM, RL-driven LLM reasoning, and LLM agent security and safety. His previous works studied the trustworthiness of DRL & DNN, and customized DNNs for security-critical applications (noisy learning, OODs). He is a recipient of multiple prestigious awards, including Google and Amazon research awards, IBM Ph.D. Fellowship and CCS Outstanding Paper Award. Going beyond academic research, he is the Head of Agent Security at Virtue AI, commercializing his research in AI agent security.
Host: ACTION Student Advisory Council