Kapil Kumar Sharma
Kapil Kumar Sharma has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Dinalan Sundaram1 shared publication
Ehab Seif Ghith
Department of Mechatronics, Faculty of Engineering, Ain Shams University, Cairo, Egypt
1 shared publication- Imam Akintomiwa Akinlade
Harvard Business School
1 shared publication - Lidia Gosy Tekeste
Eritrea Institute of Technology, Mai-Nefhi College, Himbrti, Mai Nefhi, Eritrea
1 shared publication Naveen Sri Harsha Rellu
Independent Researcher,Holly Springs,NC,USA,27540
1 shared publication- Praveena Challa
Lead Application Development Engineer, FinTech, Amazon, Seattle, USA
1 shared publication
Latest Publications
- Open Access
- 7 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
AI driven system for enhancing consumer electronics through maintenance personalization and security
The rapid expansion of consumer electronics has created an urgent need for advanced solutions to critical challenges such as predictive maintenance, user personalization, and device security. Traditional models often struggle to address these issues due to their limited adaptability and performance. This paper introduces GenAI-A, an innovative AI model that integrates Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Dynamic Recommendation Algorithms (DRA), and anomaly detection techniques to provide a comprehensive solution for consumer electronics. By leveraging the strengths of generative AI and self-renewal capabilities, GenAI-A enhances predictive maintenance, optimizes user experience, and strengthens biometric security. The model’s hybrid architecture utilizes GANs for realistic data representation, VAEs for the generation of complex data distributions, and DRA for real-time user personalization. The novelty of GenAI-A lies in its cross-regularized coupling between the GAN and VAE modules, where latent features are jointly optimized through a shared loss function to achieve consistent generative–representational learning. Unlike conventional hybrids that treat these models independently, GenAI-A introduces a dynamic feedback mechanism in which the DRA and anomaly detection modules operate directly in the shared latent space, enabling self-adaptive personalization and continual refinement of generative outputs. Experimental validation across four real-world datasets, i.e., Smartphone Sensor, Labelled Faces in the Wild (LFW), Pecan Street Energy Consumption, and SECOM Manufacturing, demonstrates significant improvements in device uptime, user engagement, and biometric security, with a notable reduction in false positives. Unlike existing hybrid generative models, GenAI-A introduces a novel integration of these components in a dynamic, self-learning system that adapts in real time to evolving user behaviors and device conditions. This unique combination of techniques sets GenAI-A apart from traditional approaches, establishing a new benchmark for AI-driven solutions in consumer electronics.
- 7 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessAI driven system for enhancing consumer electronics through maintenance personalization and security
The rapid expansion of consumer electronics has created an urgent need for advanced solutions to critical challenges such as predictive maintenance, user personalization, and device security. Traditional models often struggle to address these issues due to their limited adaptability and performance. This paper introduces GenAI-A, an innovative AI model that integrates Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Dynamic Recommendation Algorithms (DRA), and anomaly detection techniques to provide a comprehensive solution for consumer electronics. By leveraging the strengths of generative AI and self-renewal capabilities, GenAI-A enhances predictive maintenance, optimizes user experience, and strengthens biometric security. The model’s hybrid architecture utilizes GANs for realistic data representation, VAEs for the generation of complex data distributions, and DRA for real-time user personalization. The novelty of GenAI-A lies in its cross-regularized coupling between the GAN and VAE modules, where latent features are jointly optimized through a shared loss function to achieve consistent generative–representational learning. Unlike conventional hybrids that treat these models independently, GenAI-A introduces a dynamic feedback mechanism in which the DRA and anomaly detection modules operate directly in the shared latent space, enabling self-adaptive personalization and continual refinement of generative outputs. Experimental validation across four real-world datasets, i.e., Smartphone Sensor, Labelled Faces in the Wild (LFW), Pecan Street Energy Consumption, and SECOM Manufacturing, demonstrates significant improvements in device uptime, user engagement, and biometric security, with a notable reduction in false positives. Unlike existing hybrid generative models, GenAI-A introduces a novel integration of these components in a dynamic, self-learning system that adapts in real time to evolving user behaviors and device conditions. This unique combination of techniques sets GenAI-A apart from traditional approaches, establishing a new benchmark for AI-driven solutions in consumer electronics.
- 20 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Engineering Reliable Retrieval-Augmented Generation for Regulatory and Compliance Systems
Regulatory systems in domains such as healthcare, finance, and legal compliance require more than accurate question answering; they demand reliability, traceability, and stability as policies evolve over time. While Retrieval-Augmented Generation (RAG) improves factual grounding compared to standalone large language models, many deployments remain fragile in production due to retrieval mismatches, unsupported generation, and silent behavioral regressions following document updates. This paper presents a reliability-engineering approach to RAG for regulatory systems. We integrate hybrid densesparse retrieval with cross-encoder reranking, faithfulness guardrails that enforce evidence-backed generation, and continuous regression testing triggered by policy changes. Together, these mechanisms treat RAG as an evolving software system rather than a static inference pipeline. Experimental evaluation on a validated regulatory question-answer set shows that hybrid retrieval improves Recall@5 by up to 18% over dense-only baselines, while faithfulness validation reduces unsupported statements by 67%. A regression framework detects policy-induced answer changes in controlled update scenarios, supporting audit-ready operation. The proposed framework provides a practical blueprint for engineering reliable and auditable RAG systems in compliance-heavy enterprise environments.
- 20 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Engineering Reliable Retrieval-Augmented Generation for Regulatory and Compliance Systems
Regulatory systems in domains such as healthcare, finance, and legal compliance require more than accurate question answering; they demand reliability, traceability, and stability as policies evolve over time. While Retrieval-Augmented Generation (RAG) improves factual grounding compared to standalone large language models, many deployments remain fragile in production due to retrieval mismatches, unsupported generation, and silent behavioral regressions following document updates. This paper presents a reliability-engineering approach to RAG for regulatory systems. We integrate hybrid densesparse retrieval with cross-encoder reranking, faithfulness guardrails that enforce evidence-backed generation, and continuous regression testing triggered by policy changes. Together, these mechanisms treat RAG as an evolving software system rather than a static inference pipeline. Experimental evaluation on a validated regulatory question-answer set shows that hybrid retrieval improves Recall@5 by up to 18% over dense-only baselines, while faithfulness validation reduces unsupported statements by 67%. A regression framework detects policy-induced answer changes in controlled update scenarios, supporting audit-ready operation. The proposed framework provides a practical blueprint for engineering reliable and auditable RAG systems in compliance-heavy enterprise environments.
- Open Access
- 1 January 2026
- Published by Elsevier in SSRN Electronic Journal
Leveraging AI-Based Behavior Analysis for Next-Generation Automotive System Validation
Validation of modern automotive systems like ADAS (Advanced Driver-Assistance Systems) or autonomous driving systems has become a hard-to-crack nut due to the complexity of the system and huge data volumes generated during the testing phase. The classical methods that borrow heavily from predefined rules and thresholds have a hard time spotting the subtle, unexpected, or even emerging anomalous behaviors. A new framework that switches to unsupervised deep learning for anomaly detection is proposed in this paper to automate the entire validation process. By processing the data from in-vehicle networks (such as CAN, LIN, or Ethernet) in the form of multivariate time series, our proposal builds an LSTM-based autoencoder model that equips itself with the nominal system behavior without the need for labeled fault data.
- 1 January 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessLeveraging AI-Based Behavior Analysis for Next-Generation Automotive System Validation
Validation of modern automotive systems like ADAS (Advanced Driver-Assistance Systems) or autonomous driving systems has become a hard-to-crack nut due to the complexity of the system and huge data volumes generated during the testing phase. The classical methods that borrow heavily from predefined rules and thresholds have a hard time spotting the subtle, unexpected, or even emerging anomalous behaviors. A new framework that switches to unsupervised deep learning for anomaly detection is proposed in this paper to automate the entire validation process. By processing the data from in-vehicle networks (such as CAN, LIN, or Ethernet) in the form of multivariate time series, our proposal builds an LSTM-based autoencoder model that equips itself with the nominal system behavior without the need for labeled fault data.
- 1 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Cloudless Intelligence: Building Autonomous AI Agents on Edge Fabric
With the blistering development of edge computing, new horizons have opened in the field of the implementation of AI systems directly on edge devices without the need to rely on centralized cloud environments. This paper introduces the Cloudless Intelligence paradigm which is centred on creating autonomous AI agents that could operate fully within decentralized edge fabrics. We discuss the architecture blueprint of edge-native agents which utilises local computation, federated learning and lightweight inference models to support real-time decision-making, privacy friendliness and context sensing. Our suggestion of a framework combines decentralized communication technology, energy-efficient AI algorithms, and dynamic resource orchestration strategies to achieve self-guiding conduct that is independent of cloud connections. Such experimental performance characteristics on various IoT and edge use case are recorded to justify the viability and practical benefits of our cloudless agents in the aspects of latency, resilience, and energy. This study establishes the design foundation of next-generation AI systems which are self-sustaining, privacy-conscious, and are designed to run in an intermittently connected distributed environment.
- 1 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Cloudless Intelligence: Building Autonomous AI Agents on Edge Fabric
With the blistering development of edge computing, new horizons have opened in the field of the implementation of AI systems directly on edge devices without the need to rely on centralized cloud environments. This paper introduces the Cloudless Intelligence paradigm which is centred on creating autonomous AI agents that could operate fully within decentralized edge fabrics. We discuss the architecture blueprint of edge-native agents which utilises local computation, federated learning and lightweight inference models to support real-time decision-making, privacy friendliness and context sensing. Our suggestion of a framework combines decentralized communication technology, energy-efficient AI algorithms, and dynamic resource orchestration strategies to achieve self-guiding conduct that is independent of cloud connections. Such experimental performance characteristics on various IoT and edge use case are recorded to justify the viability and practical benefits of our cloudless agents in the aspects of latency, resilience, and energy. This study establishes the design foundation of next-generation AI systems which are self-sustaining, privacy-conscious, and are designed to run in an intermittently connected distributed environment.