Durga Prasad Dasepalli

Durga Prasad Dasepalli

Durga Prasad Dasepalli

Senior Technical Architect, Perficient, USA, Perficient, USA (Alumni, Cleveland State University, USA)

Title of Talk:

Decentralized Soft Computing: Orchestrating Zero-Trust Federated Learning and Explainable AI (XAI) in Enterprise Systems.


Abstract

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.


Brief Profile
Durga Prasad Dasepalli is a Senior Technical Architect at Perficient (USA) and a distinguished researcher in cloud-native architectures, MLOps, and sustainable computing. An alumnus of Cleveland State University (USA), he is the lead architect and creator of the open-source nsai-deterministic-gate framework, which specializes in neuro-symbolic guardrails to reduce computational waste in high-scale enterprise systems. With over a decade of industry-leading experience designing and deploying distributed systems for global enterprises, his academic research focuses on the intersection of federated learning, privacy-preserving AI, and explainable intelligence (XAI). Durga Prasad actively serves as a program referee and technical reviewer for numerous prestigious IEEE and Springer international conferences.