Multi-Agent Stock Prediction Systems: Machine Learning Models, Simulations, and Real-Time Trading Strategies.
Venue: (Under-Review)
This paper presents a comprehensive study on stock price prediction, leveragingadvanced machine learning (ML) and deep learning (DL) techniques to improve financial forecasting accuracy. The research evaluates the performance of various recurrent neural network (RNN) architectures, including Long Short-Term Memory (LSTM) networks, Gated Recurrent Units (GRU), and attention-based models. These models are assessed for their ability to capture complex temporal dependencies inherent in stock market data. Our findings show that attention-based models outperform other architectures, achieving the highest accuracy by capturing both short and long-term dependencies. This study contributes valuable insights into AI-driven financial forecasting, offering practical guidance for developing more accurate and efficient trading systems.

[Under-Review]

ROSTA - RObotic Sequential Task Assignment Framework for Multi-Robot Collaboration in Dynamic and Uncertain Environments.
Venue: ASCE i3CE 2025, USA (Under-Review)
Efficient task allocation is critical in multi-construction robot collaboration to enhance productivity and maximize task success rates. Robots often exhibit partial skill proficiency before researchers successfully build generalist and error-free control systems. For instance, robots are typically programmed with skills in specific activities and also may face uncertainties in execution success rates. In such scenarios, multiple robots need to coordinate sequential activities to complete construction tasks. Optimizing task allocation becomes crucial for ensuring high-quality task assignments and a high success rate in robot execution. This research presents a dynamic programming-based framework for sequential task allocation, maximizing the compatibility between task features and robot skills. Thirty-six Tasks are annotated with features such as being heavy, dexterous, or requiring careful handling. These features were robot capabilities and estimated success rate of robots handling such tasks. The framework addresses both deterministic scenarios, where success rates are fixed, and uncertain cases, where success rates are modelled as ranges and optimized using Monte Carlo simulations extracted from the TEACH dataset. Initial evaluations demonstrate significant improvements in task efficiency and success rates when compared to conventional allocation methods. By combining robotics with advanced optimization techniques, this work contributes to the development of intelligent and adaptive systems capable of solving complex task scheduling challenges in collaborative multi-robot environments.

[Under-Review]

From Assistant to Attacker: A Role-Based and Empirical Study of LLMs in Penetration Testing and Modular Architecture
Venue: ACL 2025, Austria (Under-Review)
Large Language Models (LLMs) have been explored for automating or enhancing penetration testing tasks, but their effectiveness and reliability across diverse attack phases remain open questions. In this study, we pose four research questions regarding LLMs’ roles in offensive security, their empirical performance on realistic benchmarking environments, the nature of their common failure modes, and the impact of modular versus single-agent architectures. Our experiments on Hack The Box machines and a Metasploitable setup indicate that single-agent LLMs, such as ChatGPT 4o, tend to maintain context more effectively than modular architectures, which often suffer from redundant scanning and syntax errors. Nonetheless, tasks demanding real-time analysis (e.g., man-in-the-middle attacks) showed uniform failure across all models. We conclude that while LLM-driven penetration testing can reduce manual overhead in certain subtasks, improvements are needed in error handling, multi-step coordination, and ethical oversight before these tools can be broadly relied upon in production security workflows.

[Under-Review]

Diagnostic test accuracy of AI-Assisted mammography for breast imaging: A Narrative Review
Venue: PeerJ Computer Science
The integration of artificial intelligence into healthcare, particularly in mammography, holds immense potential for improving breast cancer diagnosis. Artificial intelligence (AI), with its ability to process vast amounts of data and detect intricate patterns, offers a solution to the limitations of traditional mammography, including missed diagnoses and false positives. This review focuses on the diagnostic accuracy of AI-assisted mammography, synthesizing fundings from studies across different clinical settings and algorithms. The motivation for this research lies in addressing the need for enhanced diagnostic tools in breast cancer screening, where early detection can significantly impact patient outcomes. Although AI models have shown promising improvements in sensitivity and specificity, challenges such as algorithmic bias, interpretability, and the generalizability of models across diverse populations remain. The review concludes that while AI holds transformative potential in breast cancer screening, collaborative efforts between radiologists, AI developers, and policymakers are crucial for ensuring ethical, reliable, and inclusive integration into clinical practice.

[Paper]

AIOps-Driven Enhancement of Log Anomaly Detection in Unsupervised Scenarios
Venue: IEEE BdKCSE
Artificial intelligence operations (AIOps) play a pivotal role in identifying, mitigating, and analyzing anomalous system behaviors and alerts. However, the research landscape in this field remains limited, leaving significant gaps unexplored. This study introduces a novel hybrid framework through an innovative algorithm that incorporates an unsupervised strategy. This strategy integrates Principal Component Analysis (PCA) and Artificial Neural Networks (ANNs) and uses a custom loss function to substantially enhance the effectiveness of log anomaly detection. The proposed approach encompasses the utilization of both simulated and real-world datasets, including logs from SockShop and Hadoop Distributed File System (HDFS). The experimental results are highly promising, demonstrating significant reductions in pseudo-positives. Moreover, this strategy offers notable advantages, such as the ability to process logs in their raw, unprocessed form, and the potential for further enhancements. The successful implementation of this approach showcases a remarkable reduction in anomalous logs, thus unequivocally establishing the efficacy of the proposed methodology. Ultimately, this study makes a substantial contribution to the advancement of log anomaly detection within AIOps platforms, addressing the critical need for effective and efficient log analysis in modern and complex systems.

[Paper]

The New Frontier of Cybersecurity: Emerging Threats and Innovations
Venue: IEEE ICT
In today's digitally interconnected world, cybersecurity threats have reached unprecedented levels, presenting a pressing concern for individuals, organizations, and governments. This study employs a qualitative research approach to comprehensively examine the diverse threats of cybersecurity and their impacts across various sectors. Four primary categories of threats are identified and analyzed, encompassing malware attacks, social engineering attacks, network vulnerabilities, and data breaches. The research delves into the consequences of these threats on individuals, organizations, and society at large. The findings reveal a range of key emerging threats in cybersecurity, including advanced persistent threats, ransomware attacks, Internet of Things (IoT) vulnerabilities, and social engineering exploits. Consequently, it is evident that emerging cybersecurity threats pose substantial risks to both organizations and individuals. The sophistication and diversity of these emerging threats necessitate a multi-layered approach to cybersecurity. This approach should include robust security measures, comprehensive employee training, and regular security audits. The implications of these emerging threats are extensive, with potential consequences such as financial loss, reputational damage, and compromised personal information. This study emphasizes the importance of implementing effective measures to mitigate these threats. It highlights the significance of using strong passwords, encryption methods, and regularly updating software to bolster cyber defenses.

[Paper]

Revolutionizing Diabetic Retinopathy Diagnosis in Third World Countries: The Transformative Potential of Smartphone-Based AI
Venue: PeerJ Computer Science (Under-Review)
This narrative review explores the revolutionary impact of smartphone-based artificial intelligence (AI) in diabetic retinopathy (DR) diagnosis in third-world countries. Leveraging the widespread availability of smartphones and advanced AI algorithms, this technology offers promising solutions to overcome challenges faced by resource-limited healthcare systems. We discuss the benefits of smartphone-based AI, such as increased access to retinal screening, cost-effectiveness, timely detection, and enhanced patient engagement. Addressing challenges like image quality standardization, validation, ethical considerations, and expertise is essential for successful implementation. Smartphone-based AI has implications for healthcare delivery, including strengthened primary care, patient-centric care, and improved public health strategies. Future opportunities lie in advancements in AI algorithms, integration with wearable devices, collaborations with healthcare systems and NGOs, AI-powered disease monitoring, longitudinal data analysis, and research and development collaborations. By embracing innovation and overcoming barriers, smartphone-based AI can pave the way for a brighter future in diabetic retinopathy management and eye care delivery for all.

[Paper]

Drone-Enabled Load Management for Solar Small Cell Networks in Next-Gen Communications Optimization for Solar Small Cells
Venue: IEEE COMNETSAT
In recent years, the cellular industry has witnessed a major evolution in communication technologies. It is evident that the Next Generation of cellular networks (NGN) will play a pivotal role in the acceptance of emerging IoT applications supporting high data rates, better Quality of Service(QoS), and reduced latency. However, the deployment of NGN will introduce a power overhead on the communication infrastructure. Addressing the critical energy constraints in 5G and beyond, this study introduces an innovative load transfer method using drone-carried airborne base stations (BSs) for stable and secure power reallocation within a green micro-grid network. This method effectively manages energy deficit by transferring aerial BSs from high to low-energy cells, depending on user density and the availability of aerial BSs, optimizing power distribution in advanced cellular networks. The complexity of the proposed system is significantly lower as compared to existing power cable transmission systems currently employed in powering the BSs. Furthermore, our proposed algorithm has been shown to reduce BS power outages while requiring a minimum number of drone exchanges. We have conducted a thorough review on a real-world dataset to prove the efficacy of our proposed approach to support BS during high load demand times.

[Paper]

Improving Source-Free Target Adaptation with Vision Transformers Leveraging Domain Representation Images
Venue: arXiv
Unsupervised Domain Adaptation (UDA) methods facilitate knowledge transfer from a labeled source domain to an unlabeled target domain, navigating the obstacle of domain shift. While Convolutional Neural Networks (CNNs) are a staple in UDA, the rise of Vision Transformers (ViTs) provides new avenues for domain generalization. This paper presents an innovative method to bolster ViT performance in source-free target adaptation, beginning with an evaluation of how key, query, and value elements affect ViT outcomes. Experiments indicate that altering the key component has negligible effects on Transformer performance. Leveraging this discovery, we introduce Domain Representation Images (DRIs), feeding embeddings through the key element. DRIs act as domain-specific markers, effortlessly merging with the training regimen. To assess our method, we perform target adaptation tests on the Cross Instance DRI source-only (SO) control. We measure the efficacy of target adaptation with and without DRIs, against existing benchmarks like SHOT-B* and adaptations via CDTrans. Findings demonstrate that excluding DRIs offers limited gains over SHOT-B*, while their inclusion in the key segment boosts average precision promoting superior domain generalization. This research underscores the vital role of DRIs in enhancing ViT efficiency in UDA scenarios, setting a precedent for further domain adaptation explorations.

[Paper]

Learning Analytics and Online Courses: A Bayesian Belief Network Approach to Predict Success
Venue: Springer
In the field of educational research, learning analytics is one of the prevailing areas of exploration. The study explores a part of learning analytics using a Bayesian networks (BN) model to predict the success of the course in the online mode of education. Through the simulation results, it was found that the BN approach can be used to suggest improved online instruction delivery methods, helping the instructors and students reform their practices to maintain a synergy for a successful running of the course. As the study was executed on engineering students, it could further be generalized using students of other streams for comprehensive understanding. The study reveals that the student synergy with the method of teaching, paper difficulty, and take-home assignments are found to be the main determinants of the success of E-learning courses. The study reveals that student’s synergy with the method.

[Paper]

PotholeGuard: A Pothole Detection Approach by Point Cloud Semantic Segmentation
Venue: IEEE ICMERALDA
Pothole detection is crucial for road safety and maintenance, traditionally relying on 2D image segmentation. However, existing 3D Semantic Pothole Segmentation research often overlooks point cloud sparsity, leading to suboptimal local feature capture and segmentation accuracy. Our research presents an innovative point cloud-based pothole segmentation architecture. Our model efficiently identifies hidden features and uses a feedback mechanism to enhance local characteristics, improving feature presentation. We introduce a local relationship learning module to understand local shape relationships, enhancing structural insights. Additionally, we propose a lightweight adaptive structure for refining local point features using the K-nearest neighbor algorithm, addressing point cloud density differences and domain selection. Shared MLP Pooling is integrated to learn deep aggregation features, facilitating semantic data exploration and segmentation guidance. Extensive experiments on three public datasets confirm PotholeGuard's superior performance over state-of-the-art methods. Our approach offers a promising solution for robust and accurate 3D pothole segmentation, with applications in road maintenance and safety.

[Paper]

SAppKG: Mobile App Recommendation Using Knowledge Graph and Side Information-A Secure Framework
Venue: IEEE Access
Due to the rapid development of technology and the widespread usage of smartphones, the number of mobile applications is exponentially growing. Finding a suitable collection of apps that aligns with users’ needs and preferences can be challenging. However, mobile app recommender systems have emerged as a helpful tool in simplifying this process. But there is a drawback to employing app recommender systems. These systems need access to user data, which is a serious security violation. While users seek accurate opinions, they do not want to compromise their privacy in the process. We address this issue by developing SAppKG, an end-to-end user privacy-preserving knowledge graph architecture for mobile app recommendation based on knowledge graph models such as SAppKG-S and SAppKG-D, that utilized the interaction data and side information of app attributes. We tested the proposed model on real-world data from the Google Play app store, using precision, recall, mean absolute precision, and mean reciprocal rank. We found that the proposed model improved results on all four metrics. We also compared the proposed model to baseline models and found that it outperformed them on all four metrics.

[Paper]

SkyCharge: Deploying Unmanned Aerial Vehicles for Dynamic Load Optimization in Solar Small Cell 5G Networks
Venue: Elseiver-Computer Communications
The power requirements posed by the fifth-generation and beyond cellular networks are an important constraint in network deployment and require energy-efficient solutions. In this work, we propose a novel user load transfer approach using airborne base stations (BS) mounted on drones for reliable and secure power redistribution across the microgrid network comprising green small cell BSs. Depending on the user density and the availability of an aerial BS, the energy requirement of a cell with an energy deficit is accommodated by migrating the aerial BS from a high-energy to a low-energy cell. The proposed hybrid drone-based framework integrates long short-term memory with unique cost functions using an evolutionary neural network for drones and BSs and efficiently manages energy and load redistribution. The proposed algorithm reduces BS power outages and maintains consistent throughput stability, thereby demonstrating its capability to boost the reliability and robustness of wireless communication systems.

[Paper]

AutoAttacker
Mar 2024 - Jan 2025

AutoAttacker is an autonomous penetration testing tool powered by large language models. It leverages a RAG (Retrieval-Augmented Generation) database alongside a custom LIFT-based Masked Language Modeling framework. Using a hierarchical agent model, the system dynamically adapts to emerging threats by integrating real-time threat intelligence, advanced natural language understanding, and automated vulnerability scanning to minimize human intervention.

Blockhouse - Efficient Trading with Price Impact
Dec 2024 - Jan 2025

This project implements advanced trading algorithms based on the research paper "Efficient Trading with Price Impact." Developed in Python, it integrates statistical methods, machine learning, and high-frequency trading simulations. The codebase models both linear and nonlinear price impacts, optimizing order execution strategies and providing interactive visualizations, performance metrics, and dynamic dashboards to monitor liquidity and slippage.

CVVT
Aug 2024 - Jan 2025

CVVT combines classical computer vision techniques with modern deep learning. The project implements custom edge detection filters tuned for zebra patterns, computes epipolar geometry for stereo vision, employs robust RANSAC model fitting, and applies homography transformations for perspective correction. Additionally, a Convolutional Neural Network (CNN) trained on the CIFAR dataset demonstrates enhanced image classification through advanced data augmentation and transfer learning.

Forman-Ricci-Curvature
Nov 2024 - Jan 2025

This project provides a comprehensive Python toolkit for computing and visualizing Forman-Ricci curvature on directed, weighted graphs. By leveraging NetworkX and custom visualization methods, it analyzes edge-level geometric properties to uncover structural insights, enabling curvature-based clustering and anomaly detection in complex networks.

Image-Colouration
Aug 2024 - Jan 2025

Image-Colouration tackles the challenge of automatic image colorization using deep generative models. The project employs state-of-the-art GANs and transformer-based architectures, utilizing multi-stage training with perceptual loss and style transfer to convert grayscale images into vibrant, realistic color renditions. It is designed for both artistic enhancement and historical photo restoration.

Cleric
Sep 2024 - Nov 2024

Cleric is an innovative web application that deploys large language models in a Kubernetes environment. Built using FastAPI, Docker, and Minikube, it features automated load balancing, dynamic resource allocation, and secure microservice communications to deliver a highly scalable and fault-tolerant AI-driven service platform.

ROSTA
Aug 2024 - Jan 2025

ROSTA (Robotic Sequential Task Assignment) is a framework for multi-robot collaboration in dynamic, uncertain construction environments. It integrates sensor fusion, real-time AI planning, and adaptive scheduling algorithms to efficiently assign sequential tasks across heterogeneous robot teams, enhancing safety and operational efficiency on construction sites.

NetArch
Aug 2024 - Jan 2025

NetArch is a forward-looking project that explores next-generation network architectures. It investigates novel routing protocols, software-defined networking (SDN), and network function virtualization (NFV) to optimize performance, scalability, and security for IoT and edge computing environments.

MobAppRS
Aug 2021 - August 2023

MobAppRS is an advanced mobile app recommendation system that harnesses machine learning to analyze user behavior, preferences, and contextual data. By employing collaborative filtering, deep neural networks, and NLP techniques, it delivers personalized recommendations with real-time analytics and adaptive learning capabilities.

Stock Market Optimization
Aug 2024 - Jan 2025

This project focuses on optimizing stock market trading strategies through advanced quantitative analysis and machine learning. It incorporates historical data analysis, real-time market signals, risk management metrics, and backtesting frameworks to develop adaptive algorithms that maximize portfolio performance under volatile market conditions.

Bayesian Network E-learning
Aug 2019 - August 2022

An interactive e-learning platform dedicated to Bayesian networks and probabilistic reasoning. The system offers hands-on coding labs, interactive tutorials, real-world case studies, and adaptive learning paths to help users master Bayesian inference and its applications in data science.

GlobalBankUserManagement
Aug 2022 - Nov 2022

A robust user management system tailored for banking applications. It features advanced authentication, role-based access control, secure encryption, RESTful API integration, and a microservices architecture to ensure high scalability and regulatory compliance.

Bank
Oct 2021 - Nov 2021

A comprehensive banking system application designed for managing customer accounts. The platform features secure transaction processing, real-time balance updates, modern web frameworks, and a microservices backend—all supported by continuous integration and automated testing for high reliability.

Core Member
Enactus

Sep 2018 - May 2021 · 2 yr 8 mos

Member of the founding team of Enactus, BITS Pilani Chapter launched with the aim of social betterment through entrepreneurial measures. As a part of the exposition team, I was responsible for creating pitch decks, presenting at Regional and National competitions, and securing sponsorships. I also managed project AAVEG by overseeing financial estimates, public relations, marketing, design, publicity, and overall project management, including village visits, monitoring, and deadline coordination.

Volunteer
MyGov India

Mar 2020 - Jan 2022 · 1 yrs 10 mos

Conducting online webinars for students in rural areas, promoting social distancing measures, raising public awareness on hygiene practices, and organizing community-level campaigns across residential associations, social groups, and religious places.

Production Team Member
Hindi Drama Club, BITS Pilani

Aug 2018 - Apr 2019 · 9 mos

Ideated, created, and supported a 20-member crew to deliver stellar performances, winning the Stage Play Competition. Responsibilities included coordinating rehearsals, managing backstage logistics, and contributing to script development.

Member
Maharashtra Mandal, BITS Pilani

Aug 2018 - Aug 2022 · 4 yr

Actively participated in cultural events such as Cultural Mama Nite and Ganesh Nite. Organized event logistics, managed catering arrangements, and coordinated stage setups, ensuring engaging and memorable community events.

Campus Ambassador
IAESTE

Feb 2020 - Feb 2021 · 1 yr 1 mo

Represented IAESTE by promoting international technical internships. Worked across administration, exchange programs, business development, public relations, and finance, effectively connecting students with global opportunities.

Campus Ambassador
StartEarly

Mar 2020 - May 2020 · 3 mos

Raised awareness about StartEarly’s initiatives through targeted marketing and promotional activities. Organized workshops and seminars, led social media campaigns, and coordinated events to boost community engagement and brand visibility.

Volunteer
HelpAge India

Jun 2012 - Apr 2013 · 11 mos

Assisted in organizing health and wellness drives for senior citizens, coordinated fundraising events, and supported community outreach initiatives designed to enhance the well-being of the elderly.

Volunteer
Disha foundation

Jun 2010 - Apr 2011 · 11 mos

Contributed to community development initiatives by engaging in local outreach programs, facilitating educational workshops, and coordinating volunteer-led projects to support underprivileged communities.

Finance Raising Committee Volunteer
National Association for the Blind, India

Jun 2011 - Apr 2012 · 11 mos

Collaborated with a dedicated team to organize fundraising events and campaigns. Successfully raised financial support for the visually impaired community through effective donor engagement, event planning, and strategic outreach initiatives.

Speaker
Students'​ Academic Cell, BITS Pilani

Jan 2022 - Jun 2022 · 6 mo

Invited as a keynote speaker at BITS Pilani, where I delivered an informational talk on Practice School-II covering domains such as Software Development, Research, and Data Science.