Durga Prasad Dasepalli
Durga Prasad Dasepalli
Senior Technical Architect, Perficient, USA, Perficient, USA (Alumni, Cleveland State University, USA)
Decentralized Soft Computing: Orchestrating Zero-Trust Federated Learning and Explainable AI (XAI) in Enterprise Systems.
The rapid integration of artificial intelligence in enterprise environments has heightened the demand for highly scalable, secure, and transparent decision-making architectures. While centralized deep learning paradigms offer high predictive power, they introduce substantial vulnerabilities regarding data privacy, network latency, and black-box interpretability. This keynote address introduces a decentralized soft computing framework that integrates Zero-Trust Federated Learning (FL) with Explainable AI (XAI) to optimize complex distributed workflows. We formalize a cooperative multi-node model training protocol over localized data subsets, minimizing a global loss function without transmitting raw, sensitive data across network boundaries. To overcome the systemic distrust associated with black-box neural networks in high-stakes industries, our architecture couples this decentralized optimization with local SHAP (Shapley Additive exPlanations) and LIME-based feature attribution models. We demonstrate how combining lightweight symbolic constraints with stochastic models, the neuro-symbolic approach, drastically slashes active GPU runtime by up to 48% and guarantees absolute compliance with strict zero-trust parameters. Attendees will gain actionable insights into deploying scalable, privacy-preserving MLOps pipelines in heterogeneous multi-cloud environments.