Computer Engineering, Computer Networks and Communications, Agricultural and Biological Sciences, Management Information Systems
20
Scopus Publications
Scopus Publications
A Cloud-based LLM-Powered RAG and SQL Agent Framework for Reliable Personal Finance Assistance L Kartheesan, Syed Faheez Ahamad, Muvvala Sai Puneeth Proceedings of 6th International Conference on Expert Clouds and Applications Icoeca 2026, 2026 Accelerated advancements in Large Language Models (LLMs) has enable smarter financial assistant systems. This paper presents an in-depth development roadmap for a cloud-native personal finance chatbot that will be built upon Retrieval-Augmented Generation (RAG) in tandem with SQL agents to provide accurate and financially relevant guidance. The envisioned system aims to address the existing critical challenges of AI hallucination, data staleness, and the need for personalized financial advice, by incorporating external knowledge retrieval as well as structured database querying. It explore the theoretical underpinnings of RAG systems, analyze multi-agent SQL generation techniques and propose an architecture that unifies these technologies to build a robust personal finance assistant. Consequently, the system will be able to handle complex financial queries, while also ensuring production-grade scalability and reliability through a centralized cloud-based reasoning engine.
Energy-aware cluster head optimization and secure blockchain integration for heterogeneous 6G-enabled IoMT networks R. Yuvarani, R. Mahaveerakannan, Tamilvizhi Thanarajan, L. Kartheesan Scientific Reports, 2025 The integration of blockchain into 6G-enabled Internet of Medical Things (IoMT) networks promises secure and decentralized communication but introduces challenges related to energy efficiency, latency, and authentication overhead. Existing clustering and security schemes fail to balance these aspects effectively in heterogeneous networks. This paper proposes a novel energy-aware cluster head (CH) selection framework using Artificial Democratic Cuckoo Glowworm Remora Optimization (ADCGRO), integrated with a lightweight blockchain layer for secure authentication and data integrity. The system optimizes task allocation across advanced, intermediate, and normal IoMT devices to minimize energy depletion while meeting ultra-reliable low-latency communication (URLLC) requirements. Simulation results demonstrate that the proposed approach enhances network lifetime by 27%, reduces average latency by 35%, and achieves 99% authentication accuracy, surpassing baseline protocols such as LEACH and HEED. These results highlight the effectiveness of combining ADCGRO-based optimization with blockchain to enhance performance and security in 6G wireless networks.
Energy-Efficient Cluster-Based Reliable Routing Using Hybrid Nutcracker and Improved Sand Cat Optimization Algorithm for Extending Network Lifetime in WSNs Joseph Martin Sahayaraj, Gopi Prabaharan, Loganathan Kartheesan, Natarajan Jayapandian International Journal of Communication Systems, 2025 In wireless sensor networks (WSNs), sensor nodes are deployed in a target region for sensing environmental physical parameters to attain the objective of reactive decision‐making. These sensor nodes necessitate energy for processing and forwarding the sensed data to the base station (BS) for better data delivery in WSNs. Balanced energy utilization in WSNs prevents the problem of hotspot, and dynamic cluster head (CH) selection with reliable route establishment is a vital decision‐making approach that helps in optimal path selection with maximized energy conservation. In this paper, a nutcracker and sand cat optimization algorithm (NCSCOA)–based multiobjective CH selection and sink node mobility scheme is propounded for enabling rapid and reliable data transmission with reduced energy consumption in heterogeneous WSNs. This NCSCOA handled the problem of hotspot as well as isolated nodes and facilitated loop‐free routing with the support of the improved nutcracker optimization algorithm (INCOA) that makes the decision of routing using local and global search optimization processes. It constructed an energy‐level matrix (ELM) by deriving the impactful factors of intercluster formation, distance between CH and BS, residual energy (RE), and node density for achieving optimal CH selection and route determination. In specific, improved sand cat optimization algorithm (ISCOA) is used during the intercluster formation phase by discovering the optimized path between source and destination during route establishment. Simulation‐based findings of the proposed NCSCOA confirmed its efficacy by improving the mean number of alive nodes by 23.18%, reducing energy consumption and delay by 21.86% and 20.98% compared to benchmarked protocols.
Lung disease classification in chest X-ray images using optimal cross stage partial bidirectional long short term memory T Babu, GV Sam Kumar, L Kartheesan, Surendran Rajendran Journal of X Ray Science and Technology, 2025 Background Lung disease is the crucial disease that affects the breathing conditions and even causes death. There are various approaches for the lung disease classification; still the inefficiency in accurate detection, computational complexity and over-fitting issues limits the performance of the model. To overcome the challenges, a deep learning model is proposed in this research. Initially, the input is acquired and is pre-processed using three various techniques like data augmentation, filtering and image re-sizing. Then, the threshold based segmentation is employed for obtaining the required region. Objective From the segmented image, various categories of lung diseases like COVID, lung Opacity, Pneumonia and normal are identified using the proposed Optimal Cross Stage Partial Bidirectional Long short term memory (OCBiNet). Methods The proposed OCBiNet is designed using Bidirectional Long short-term memory (BiNet) with Cross Stage Partial connection in its hidden state. Besides, the adjustable parameters are modified using the proposed Improved Mother Optimization (ImMO) algorithm. Results The ImMO algorithm is designed by integrating the Logistic Chaotic Mapping within the conventional Mother Optimization algorithm for enhancing the convergence rate in obtaining the global best solution. Conclusions The proposed OCBiNet is evaluated based on Accuracy, Recall, Precision, and F-Score and acquired the values of 99.11%, 98.98%, 99.18%, and 99.08% respectively.
Lexicon-Enhanced Convolutional BERT (LEConvBERT) for Secure and Intelligent Equivalency Certificate Generation using NLP and Blockchain Sumathy Krishnan, Surendran R, Sangeetha S, Kartheesan L Proceedings of 3rd International Conference on Sustainable Computing and Data Communication Systems Icscds 2025, 2025 The academic credentials and course equivalency checking across learning institutions and international borders is a challenging process that's lengthy and cumbersome, especially against an imagined student mobility and remote learning trend. Nations of assessing the equivalency of academic transcripts using established manual procedures are not always consistent or transparent, as well as prone to errors or fraud. In order to overcome these problems, this paper proposes an innovative AI-based system based on a combination of Lexicon-Enhanced Convolutional BERT (LEConvBERT) and blockchain technology to support automated, transparent, and tamper-proof generation of equivalency labels. LEConvBERT is an enhanced version of the traditional BERT-based models and introduces domain-specific lexicons and convolutional filters, which employ a broader scope of refined semantic and syntactic patterns topical to academic terms and institutional nomenclature. This helps in mapping course outcomes, credit hours, and curriculum goals better in various educational systems with better accuracy. The proposed system, which uses LEConvBERT in a natural language processing (NLP) pipeline, automatically compares academic records to understand their meaning and checks if they are equivalent, then it starts issuing certificates as smart contracts on a blockchain ledger. This system also depicts low-latency certificate creation, which makes its application ideal in real-time applications in academically transferring the information in employment screening and in admitting people internationally and stands as a scalable and complete solution to a decades-old challenge in educational technology and as an entrée into the age of intelligent, decentralized academic credentialing.
Multi-Layered IoT Biopackaging with RTD/NFC: A Printed Electronics Approach to Sustainable Logistics L Kartheesan, B Rajakumar, S Nagarajan, R Surendran Proceedings of the 6th International Conference on Electronics and Sustainable Communication Systems Icesc 2025, 2025 The fast-paced nature of the work in logistics due to e-commerce’s growth has, however, made an environmental assault because it depends majorly on plastics that are not biodegradable. This research examines the possible use of biodegradable materials like PLA, PHA, and cellulose with printed IoT sensors. It integrated such a packaging concept into one package design that is sturdy and compostable. Currently, it is able to measure temperature, humidity, and tampering events in real-time using RTD sensors, capacitive humidity sensors, and BLE/NFC antennas. These sensors are printed on environmentally friendly, flexible disposable substrates employing scalable technologies. It comprises a PLA base for mechanical strength, the printed electronics embedded for functionalities, and beeswax-shellac coating on the exterior to prevent moisture ingress too fast. Composting tests $\left(58^{\circ} \mathrm{C}, 50 \% \mathrm{RH}\right)$ revealed that PHA could degrade as much as 72% in 30 days. Sensor response continued beyond 25 days when the sensor displayed the capability to continue communication with the BLE app with a more than 95% successful data transmission rate. This packaging concept goes beyond current smart packaging limitations in providing one common solution embracing sustainability and scalability for cold-chain logistics, high-value shipments, and perishables. It offers the foundation in guiding Green Chemistry, Sustainable Electronics, and Smart Logistic integration into next-generation e-commerce packaging.
Quantum-Driven Agricultural Innovation: Evaluating Wheat Flour Quality Through Thermal Imaging and Biophotonic Emissions Shoba. B, Kartheesan. L, Priyanka. S A, Deepa. R, Surendran. R 3rd IEEE International Conference on Data Science and Network Security Icdsns 2025, 2025 Crop quality and productivity are highly dependent on advancements in agricultural innovation. Quantum methods are a new strategy to maximize plant growth and food quality. Despite positive results, several issues remain. Establishing wave-based measuring standards, ensuring environmental adaptation, and validating results in various soil conditions are these problems. Using Quantum Wave-Assisted Thermal Biophotonic Analysis (QW-TBA), this research evaluates wheat flour quality. This method involves quantum theory, thermal imaging, and biophotonic emission analysis. QW-TBA encodes each thermal imaging pixel with wave-based parameters, specifically frequency-domain components and amplitude variations derived from biophotonic emission signals, enabling a spatially resolved spectral analysis of crop tissue responses. This technology connects biophotonic emissions to flour quality parameters to create non-invasive, real-time agricultural monitoring systems. The initial data suggests that this agronomic strategy might improve crop quality assessment accuracy.
Advanced OptiDLCardioNet-Based Cardiac Arrhythmia Detection Model from ECG Signals Muthukumar B, Kartheesan L, Vijayalakshmi Pasupathy, Surendran R 5th International Conference on Sustainable Communication Networks and Application Icscna 2024 Proceedings, 2024 Heart disease is regarded as one of the most significant health problems dealing by people today. Approximately, cardiovascular disease (CVD) affects around 50 million people. Electrocardiogram (ECG) signals are more vital for identifying and keeping track of individuals with various CVDs. In order to detect different types of arrhythmias, this papers proposed a novel optimization driven deep learning model for cardiac arrhythmia detection, termed as OptiDLCardioNet. To improve and smoothen the ECG signal, a cascaded wavelet augmented Kalman (CWAK) filtering approach is first applied. Next, an Adaptive Position aware Black-winged SqueezeNet (APBWSqueezeNet) model is used for feature extraction. In order to classify the signals for arrhythmia disease diagnosis, the extracted features are input into an Enhanced Dilated Height-Width Axial Attention Convolutional Network (EDilBW-HWAACNet). Moreover, the hyper-parameters of the EDilBW-HWAACNet are adjusted through the application of the Improved Walrus Optimization Algorithm (IWOA). The MIT-BIH arrhythmia database is used to validate the performance of OptiDLCardioNet model. Rendering to the experimental results, the OptiDLCardioNet model is able to achieve high classification accuracy of 99.82%, which is superior to existing methods with fewer significant features.
Multi-Task Distillation Learning for Coffee Corticium Salmonicolor Pink Berry Disease for Real-Time Prediction Raveena S, Surendran R, Sangeetha M, Kartheesan L 5th International Conference on Sustainable Communication Networks and Application Icscna 2024 Proceedings, 2024 Coffee Corticium Salmonicolor Pink Berry Disease (CCSPBD) is a substantial risk to coffee cultivation. Precise and prompt identification is essential for efficient disease treatment. This research introduces a multi-task distillation learning (MTDL) method for the real-time prediction of CCSPBD. A more extensive teacher model is trained on a comprehensive dataset of coffee plant diseases, and its knowledge is conveyed to a smaller student model tailored for CCSPBD prediction. Integrating supplementary activities about coffee plant health into the teacher model enhances the student model's comprehension of plant disease patterns. The resultant student model can precisely forecast CCSPBD in real-time, allowing farmers to implement prompt interventions to avert crop loss. Experimental findings indicate the efficacy of the suggested MTDL method in attaining high accuracy and efficiency for CCSPBD prediction.
