Agent Systems for Cyber Defense -- Talk with Abnormal AI

​As AI-generated attacks rise, the traditional ML stack is becoming too slow and human-dependent to keep up.

​In this session, we explore how Abnormal AI has re-architected the modern ML stack by moving from pure probability inference to agentic reasoning. We will dive into how specialized agents are used not just for classification, but to automate the entire lifecycle of a defense system: from online detection using reasoning models to offline rule synthesis and automated root-cause analysis.

​We will deconstruct the architecture of an "AI Security Analyst," examining how to build stateful, tool-using agents that can perform sandboxed code execution, recursive deep research, and self-correction.

---
Interested in learning more or working at Abnormal AI?
- https://blog.sshh.io/ (Shrivu's AI blog)
- https://abnormal.ai/careers/jobs/7593366003 (AI x DevOps)