Publications

  • Home
  • Academics
  • Computer Science and Engineering (DS)
# Published Date Details
1.24-02-2026Biomedical data such as ECG and EEG are typically infrequent, private, or class-unbalanced, thereby posing extremely difficult conditions to develop strong machine learning models. To alleviate these issues, this work proposes a new synthetic data generation and classification paradigm integrating physics-constrained Generative Adversarial Networks (GANs) with a deep hybrid Convolutional Neural Network (CNN) and Transformer framework. The physics-constrained GANs are tuned to generate physiologically viable and realistic biomedical signals and can be applied for data augmentation in privacy-sensitive environments. The generated artificial data is used for training a strong model for classification that can provide disease profiling under different conditions. There is extensive validation on benchmark and semi-real datasets to assess the generalizability and performance of the proposed approach as a whole. Results show that the framework is good at classification, even with limited data, and highlights its applicability for real-world clinical use where data availability and confidentiality are of paramount importance.
Authors: Pramath KP; Banupriya Mohan; Kalyan Nagaraj; Shambhavi BR
Link
2.21-01-2026Emerging applications in smart healthcare, intelligent transportation, autonomous robotics, and cybersecurity necessitate AI systems that extend beyond traditional automation to exhibit agentic capabilities-namely, autonomy, goal-directed behavior, and contextual awareness. This paper introduces a novel hybrid Agentic AI framework that combines Artificial Intelligence Markup Language (AIML) for symbolic reasoning with data-driven machine learning models for adaptive policy learning. The proposed architecture leverages AIML’s rulebased structure to enable semantic interpretation and highlevel decision logic, while ML modules provide real-time environmental perception and dynamic strategy adjustment. The system architecture supports continuous adaptation to changing operational contexts and alignment with long-term objectives. Experimental evaluations conducted across multiple domains, including autonomous navigation and intelligent virtual assistants, demonstrate superior performance in decision accuracy, adaptability, and contextual responsiveness compared to conventional AI baselines. The results underscore the effectiveness of hybrid symbolic-subsymbolic integration in building scalable, interpretable, and ethically grounded intelligent agents. Future work will focus on expanding the framework for multi-agent coordination and human-AI collaborative systems.
Authors: Priyanka Shivaramaiah, Roopesh Ramesh, Kalyan Nagraj
Link
3.01-01-2026“barfLM: A Small Language Model” has been published in the 6th International Conference on IoT Based Control Networks and Intelligent Systems (ICICNIS) | 979-8-3315-1520-1/25/$31.00 ©2025 IEEE | DOI: 10.1109/ICICNIS66685.2025.11315709
Authors: Nayana N. Kumar
Link
4.01-01-2026Title: Advances in Detecting Deepfake Threats, Methods and Societal Implications. Published in 2025 6th International Conference on IoT Based Control Networks and Intelligent Systems (ICICNIS) ISBN: 979-8-3315-1520-1 Publisher: IEEE
Authors: Navya Damodar, Asha S Manek
Link
5.30-12-2025"Enhancing Educational Applications with Retrieval-Augmented Generation, Question Generation and Summarization Systems"
Authors: Stuthi Shrisha; Sindhu K; Sahithi Kandula; Suhas Shenoy Udyavara
Link
6.19-12-2025Managing personal finances in today's fragmented fintech landscape presents significant challenges due to the absence of integrated, user-aware systems. This work proposes a unified, intelligent finance management framework that automates budgeting, credit recommendation, risk detection, and interactive advisory services. The system combines modular components-namely, a fraud detection engine using XGBoost, a personalized credit card ranker powered by learning-to-rank models, and a finance chatbot built on LLM with intent and context detection pipelines. Key innovations include categoryaware document embeddings, semantic user profiling, and rulebased behavior mapping for spending trends. Evaluation results demonstrate effective fraud classification, accurate credit matching, and scalable document-grounded advisory generation. The platform is designed for young professionals, freelancers, and digital users, and emphasizes scalability, explainability, and contextual intelligence across financial tasks.
Authors: Pavithra Ganta; Gourav Agarwal; Rajshekhar Jha; Kalyan Nagaraj
Link
7.19-12-2025Voice assistants have become essential tools for everyday tasks, yet their inability to provide domain-specific expertise, emotional awareness, and personalized interactions limits their effectiveness in specialized contexts such as mental health, finance, and education. This paper introduces Aurora, a next-generation AI voice assistant designed to address these gaps through a multi-personality architecture. Aurora integrates distinct personas such as a therapist, financial advisor, doctor and teacher-each with tailored knowledge domains, tones, and behaviors, enabling dynamic personality switching based on user's queries. The system employs advanced text-to-speech synthesis to adapt tone, pitch, and speaking style appropriate to each persona and emotional context, while a toxicity detection module ensures safe and ethical interactions by filtering inappropriate inputs. Built with technologies like Whisper for speech recognition, Eleven labs for expressive voice synthesis, and a session-based memory for contextual recall, Aurora delivers natural, human-like conversations. Preliminary evaluations highlight its potential to enhance user engagement and trust in specialized domains, paving the way for more intelligent, adaptive, and responsible voice technologies.
Authors: Bommireddy Neha; G Sri Sai Meghana; Meet Jain; Kalyan Nagaraj
Link
8.19-12-2025"Edge-AI Enabled Real-Time ECG and Vital Sign Monitoring System for Elderly Patients"
Authors: Manoj R; Skandan A; Hemanth Umashankar; Sindhu K
Link
9.21-11-2025Building Trust in AI: A Hybrid Approach to Combating Fake News and Misinformation
Authors: Piyush Verma; Manoj Kumar S; Harsh Deep; Kumar Chinmay
Link
10.05-07-2025The research work SleepPred: A machine learning tool to predict quality of sleep was presented at the International Conference on Data-Processing and Networking (ICDPN-2024) on 25th – 26th October 2024 organized by the Institute of Technology and Business (VŠTE) in , Czech Republic, Europe.
Authors: Kalyan Nagaraj, Amulyashree S
Link
11.27-06-2025IntelliMotion: Real-Time Hand Gesture and Pose Tracking Using OpenCV and MediaPipe
Authors: Harsh Deep; Manoj Kumar S; Piyush Verma; Kumar Chinmay
Link
12.04-09-2024The research work "A Novel ensemble model for prediction of occurrence of cancer" was presented on 5th Congress on Intelligent Systems international conference on 4th – 5th September 2024 organized by Christ University.
Authors: Kalyan Nagaraj, Prashanth H S, Amulyashree S
Link
13.28-11-2023"Adaptive Proximity Alert and Currency Detection", International Conference on the Confluence of Advancements in Robotics, Vision and Interdisciplinary Technology Management (IC-RVITM), Bangalore, India, November 2023, pp. 1-6, doi: 10.1109/IC-RVITM60032.2023.10435290.
Authors: Sindhu K, C. Manjeshwar, S. Koushik, K. K. Potta
Link
14.28-11-2023"3D Learning Experience Using Augmented Reality", International Conference on the Confluence of Advancements in Robotics, Vision and Interdisciplinary Technology Management (IC-RVITM), Bangalore, India, November 2023, pp. 1-6, doi: 10.1109/IC-RVITM60032.2023.10435321.
Authors: Sindhu K, Rajath R, R. Kashyap, C. Roopesh Reddy
Link
15.01-09-2023Title: Deep Learning Models for Stock Market Prediction Using Optimization Approach, 2023 IEEE International Conference on Network, Multimedia and Information Technology (NMITCON)
Authors: BL Shilpa, BR Shambhavi
16.17-12-2021Title: Sentiment Classification on Bilingual Code-Mixed Texts for Dravidian Languages using Machine Learning Methods Source: Working Notes of FIRE 2021 - Forum for Information Retrieval Evaluation
Authors: Rashmi K.B., Guruprasad H.S., Shambhavi B.R
Link
17.06-11-2020“Demand Forecasting for E-Commerce Platforms”, IEEE International Conference for Innovation in Technology (INOCON), 06-08th November 2020, Nagarjuna College of Engineering and Technology, Bangalore
Authors: Sindhu K, Anupriya, Vikram Karthikeyan, B. Sahana, B. R. Shambhavi, S. Balaji.
Link
18.31-10-2020 Title: Recognition of Named Entities and Categories in Text using Stacked Embeddings 2020 IEEE 5th International Conference on Computing Communication and Automation (ICCCA), 2020, pp. 95-101 DOI: 10.1109/ICCCA49541.2020.9250886.
Authors: P. Narendranath, H. Jayanthi, N. Shreyas, B. Shambhavi, P. Jayarekha and J. Nabi
Link
19.07-10-2020"Recommendations in Social Network using Link Prediction Technique", IEEE , 2020 International Conference on Smart Electronics and Communication (ICOSEC), DOI: 10.1109/ICOSEC49089.2020.9215236
Authors: BV Ramya, Sandeep Varma N and Indra R
Link
20.01-09-2020"Real-time Conversion of Sign Language to Text and Speech", 2020 Second International Conference on Inventive Research in Computing Applications (ICIRCA), DOI: 10.1109/ICIRCA48905.2020.9182877, IEEE Xplore.
Authors: Kohsheen Tiku, Jayshree Maloo, Aishwarya Ramesh, Indra R
Link