Venkata Kancherla is a cloud security and AI-driven DevOps researcher specializing in Zero Trust architecture, federated learning, and securing MLOps pipelines. He is Senior Devops Consultant at Capgemini US, with 15+ years of experience in Kubernetes  and cloud security.

Title: Federated Anomaly Detection in Kubernetes-Orchestrated MLOps Pipelines: A Zero Trust Approach to Securing Distributed AI Workloads -    Abstract: MLOps pipelines running on Kubernetes and multi-cloud infrastructure have become a major, under-defended attack surface — exposed to data poisoning, model exfiltration, and CI/CD tampering that centralized security tools miss. This talk presents a federated anomaly detection framework that combines federated learning with Zero Trust to secure distributed AI workloads without centralizing sensitive data. Each cluster trains a local detection agent and shares only encrypted gradient updates, which a federated aggregator combines via Federated Averaging into a global threat model redistributed to every node. Against a centralized baseline, the approach reaches 96.3% detection accuracy, cuts latency by roughly 78%, halves false positives, and preserves data locality for GDPR/HIPAA compliance. The session also covers key limitations and future directions, including generative-AI rule synthesis and quantum-resilient channels.