Python, AI Agents, and the Future of Infrastructure Automation

Are our infrastructure platforms ready for AI-powered workflows? As AI agents gain the ability to automate deployments, generate infrastructure code, access cloud resources, and interact with production systems, traditional security policies are no longer sufficient. Through real-world examples from cloud and high-performance computing (HPC) environments, this talk explores the practical challenges of AI-enabled infrastructure and the guardrails needed to use it safely.

We'll examine risks such as credential exposure, excessive automation, resource abuse, and uncontrolled infrastructure changes, before demonstrating practical engineering approaches including least-privilege access, human approval workflows, auditing, and infrastructure guardrails. Attendees will leave with actionable recommendations for using AI safely in Python-based workflows, along with practical security best practices for both infrastructure maintainers and everyday developers.