Volume 9 Issue 4
Author's Name: N Mahaviraswamy ,Chandana V , Jyothsna I , Kavana P , Kavya B J
Abstract—Abstract— Cloud Computing is the new field that was invented and developed during a period not so long ago. It has a lot of benefit such as decreasing the cost that the user needs, process the operation faster and keep the information secure. The security of cloud is the most important issue for several sensitive occupations, for web base information using this kind of computing does not need more than a computer and high speed internet to use application which is developed by cloud computing. This paper presents a model of web based information system at Amazon Web Services using visual studio to build a website with cloud data service which takes data from the instance in EC2 on Amazon Web Services (AWS), then lets AWS host data services from the cloud. The environment used, is EC2 (Elastic Cloud Computing) on AWS (Amazon Web Services) as host for an application which has several services to allow each user use the application separately and securely. AWS configuration and the way of using virtual server which its given by EC2 will be pointed out in this paper..
Author's Name: J.R.Arun Kumar, Karan Singh, Bhawana Choudhary, Jinesh Jain, Disha Faujdar,
Abstract— Software development is an iterative process where continuous learning, experimentation, and problem-solving shape a developerʼs growth. While many developers maintain daily notes or activity logs, these records are often fragmented and lack structure, making it difficult to extract meaningful insights. As a result, valuable personal knowledge remains unused, and opportunities for reflection or self-improvement are often missed. “Dev Journey through AI — Smart Dev Log Insights” addresses this challenge by leveraging Artificial Intelligence and Natural Language Processing to transform unstructured developer logs into meaningful summaries and insights. The system automatically analyzes journal entries to extract skills, keywords, sentiments, and progress trends, enabling developers to gain a deeper understanding of their work patterns. It also generates contentready drafts suitable for sharing on social platforms, bridging the gap between personal productivity and public presence
Author's Name: Mohit Sharma, Ajay Kumar Saini, Amarnath Singh,Aryan Meena, Amrit Kumar Prajapat
Abstract— Access to efficient weed management remains a major challenge in modern agriculture, especially for small-scale farmers. Manual inspection is labor-intensive and time-consuming, while excessive use of herbicides leads to environmental damage and increased costs. This project proposes a deep learning-based weed detection system that uses computer vision to automatically classify crops and weeds from images. A Convolutional Neural Network (CNN) extracts features such as shape, texture, and color to ensure accurate classification. The system supports real-time image capture and analysis, improving early weed detection and reducing human effort. Advanced techniques like object detection can further enhance precision by locating weeds in fields. Deployable as a user-friendly web or mobile application, the system improves efficiency, reduces chemical usage, and supports sustainable farming practices, contributing to smarter agriculture and improved crop productivity worldwide.
Author's Name: Sumit Kumar, Devesh kumar, Gaurav Goel
Abstract—Abstract— Cloud computing describes effective computing services provided by a third-party organization known as cloud service provider for organizations to perform different tasks over the internet for a fee. Cloud service provider’s computing resources are dynamically reallocated per demand, and their infrastructure, platform, and software, and other resources are shared by multiple corporate and private clients. With the steady increase in the number of cloud computing subscribers of these shared resources over the years, security on the cloud is a growing concern. In this review paper, the current cloud security issues and practices are described and a few innovative solutions are proposed that can help improve cloud computing security in the future areas.
Author's Name: Mohit Sharma, Priyanshu Redu, Nilesh Singh Chouhan,Rajat Gupta, Sagar Saini
Abstract— Medical Image Analysis Systems (MIAS) are critical for automated diagnosis and clinical decision support. These systems use advanced image processing techniques to detect, segment, and classify medical images from modalities such as MRI, CT, X-ray, and Ultrasound. This paper provides an overview of key methodologies used in medical image analysis, including image pre-processing, segmentation, feature extraction, and classification using machine learning and deep learning techniques. The study discusses challenges such as noise reduction, accurate boundary detection, and computational efficiency. Applications in tumor detection, organ segmentation, and disease monitoring are highlighted. The paper also examines the advantages of automated systems over manual analysis, emphasizing improved accuracy, reproducibility, and speed. Future research directions include integrating AI with real-time imaging and predictive analytics.
Author's Name: Kalyani A, Barath P, Dinesh E, Gowthamaraj C, Kathiresan L
Abstract—Abstract— Pedestrian loads that may cause excessive structural vibration involve some uncertain parameters such as walking frequency, step length, dynamic load factors and phases of harmonic components, which will lead to uncertainties of structural response and this issue need to be solved by probabilistic analysis. Considering that the traditional Monte Carlo simulation method for reliability analysis has rather low efficiency, an approach based on uniform design and response surface method for calculating the probabilistic structural response induced by pedestrian vertical loads is proposed to improve the efficiency of structural dynamic analysis with uncertainties. A few representative samples of time history of pedestrian loads are simulated using uniform design first, and then the corresponding peak acceleration response spectra are obtained by dynamic analysis on beam structures with different spans and damping ratios. The spectra which have a certain percentile are obtained by reliability analysis based on response surface method. Then the general formulae of peak acceleration response spectra, which can be used to calculate structural peak accelerations directly, are deduced from parametric analysis of damping ratio and span. Monte Carlo simulation is conducted to validate the precision of this method. The case study shows that compare to the results calculated by the proposed method, the formulae in two widely-used codes such as BS 5400-2:2006, overestimate the peak acceleration of structure with high frequency remarkably and it should be cautious when using them to obtain structural responses..
Author's Name: Arvind Sharma, Utkarsh Gupta, Nikhil Choudhary, Praveen, Ansh Jaiswal, Sandeep Saini
Abstract— Speech Emotion Recognition (SER) is an emerging area in affective computing that enables machines to detect human emotions from speech signals. This project presents a deep learning-based SER system capable of classifying emotions such as happiness, sadness, anger, fear, and neutrality. The system extracts acoustic features including Mel-Frequency Cepstral Coefficients (MFCCs), chroma features, spectral contrast, pitch, and energy from speech signals. These features are used to train machine learning and deep learning models such as Support Vector Machines (SVM), Random Forest, and a hybrid CNN-LSTM architecture .Experimental results demonstrate that deep learning models outperform traditional methods due to their ability to capture complex temporal and spectral patterns. The system supports real-time emotion recognition and can be applied in domains such as human-computer interaction, healthcare monitoring, and smart assistants.
Author's Name: Kalyani A, Goudhamaram R, RohanSreenath E N, SuriyaPrakashR, VetrivelT
Abstract—Abstract— This study aims at testing and comparing the tensile strength of bamboo and steel reinforcement bars as structural material for building construction. Tensile strength tests were carried out on various sizes steel and bamboo; categories of reinforcement bars such as; 10mm, 12mm, 16mm, 20mm and 25mm of both high-yield and mild-yield steel reinforcement bars were both tested along with same sizes of bamboo with 10mm cross-sectional thickness. Results are presented in tables and graphs and show that the tensile strength of high-yield steel bars outstrips that of mild-yield and bamboo respectively.
Author's Name: Neeraj Jain, Tushar Yadav,Aditya Kumar,Chandra Prakash,Muskan Jangir
Abstract— Road accidents remain one of the leading global causes of death, with millions of victims affected annually due to delayed detection and slow emergency response. Traditional accident reporting systems rely heavily on eyewitness accounts, manual CCTV monitoring, and human-driven alerts, all of which are prone to fatigue, inaccuracy, and significant delays. To address these limitations, this research proposes an AI-powered real- time road accident detection system that integrates computer vision, deep learning, and automated alerting frameworks.The system analyzes live video streams from CCTV cameras, dashcams, and roadside units through a combination of YOLO-based object detection, CNN feature extraction, and motion pattern analysis. By detecting collision events, abnormal motion behavior, and sudden vehicle deceleration, the model can classify accident scenarios with high accuracy. A backend module automatically generates alerts with location, timestamps, and visual evidence, enabling faster emergency response.
Author's Name: Mohit Sharma, Umesh Kumar, Naveen, Rudraksh,Punit Sangwan
Abstract—The rapid growth of digital technology has transformed the traditional education system into a more flexible and accessible learning environment. This project presents an advanced E-Learning Platform integrated with an AI-based Quiz Generator designed to enhance the learning experience for students and educators. The platform provides users with access to study materials, video lectures, and interactive content anytime and anywhere. A key feature of this system is the implementation of Artificial Intelligence to automatically generate quizzes based on the learning content. The AI analyzes the provided study material and creates relevant questions, including multiple-choice, true/false, and short-answer formats. This reduces the manual effort required by educators and ensures continuous assessment of students.
Author's Name: J.R. Arun Kumar, Harsh Kumar, Anshul Singh, Akshat Jain
Abstract—Abstract— The growing need for emotionally intelligent and linguistically inclusive conversational systems has driven the development of an Emotion-Aware Multilingual Conversational AI with Code-Switching Support. This paper presents a comprehensive AI-powered chatbot platform capable of detecting user emotions, understanding mixed-language input such as Hinglish, and generating empathetic, contextually appropriate responses in real time. The system integrates transformer-based models including BERT and MuRIL for multilingual emotion recognition, IndicTrans2 for culturally sensitive translation, and DialoGPT or BlenderBot for coherent multi-turn dialogue generation. A modular pipeline combines preprocessing, language identification, emotion detection, code-switch- aware encoding, retrieval-augmented generation, and post-processing to deliver responses that reflect both the linguistic preference and emotional state of the user. The frontend is built using React, while Flask powers the backend API, with SQLite and MongoDB handling data storage across sessions. The system addresses critical limitations of existing conversational agents, including emotional insensitivity, poor support for code-switched communication, lack of cultural adaptability, and ineffective handling of high-sensitivity domains such as mental health and education. Results demonstrate the platform's capability to process multilingual and code-mixed conversations with emotional awareness, laying a strong foundation for inclusive, human-centered AI communication systems applicable across diverse global populations.
Author's Name: J.R.Arun Kumar,Aman Sahu, Kartik Jain, Aryan Khandelwal,Kunal
Abstract—Abstract— MediPredict is an AI-powered healthcare application designed to detect and analyze skin diseases at an early stage using image processing and deep learning techniques. Skin diseases are common worldwide, yet timely diagnosis remains a challenge due to lack of awareness, accessibility issues, and dependency on medical professionals. This system leverages Convolutional Neural Networks (CNN) to analyze uploaded skin images and predict possible diseases such as acne, psoriasis, melanoma, and fungal infections with high accuracy. The model extracts visual features like texture, color, and patterns to classify conditions and generate confidence scores. Additionally, MediPredict provides structured guidance including symptoms, causes, prevention methods, and recommendations for medical consultation. The system integrates a mobile application (Flutter), cloud backend (Firebase), and AI models for real-time predictions. Experimental results demonstrate that MediPredict offers reliable preliminary diagnosis and improves healthcare accessibility. It acts as a supportive tool for early detection and awareness, contributing toward digital healthcare transformation.
Author's Name: : Mohit Sharma, Mohit Pradhan, Harsh Gupta, Abhishek Pandey, Dinesh Kumar, Ajay Jatav
Abstract— The ML-Powered Book Recommendation System provides personalized book suggestions by analyzing user preferences and past interactions using content-based filtering. It evaluates features such as genre, author, keywords, and descriptions to identify patterns and recommend books aligned with user interests. This approach enhances the reading experience by simplifying the discovery of relevant content and encouraging exploration of new topics. The system reduces the effort of browsing large collections by automatically filtering suitable options. It incorporates key processes like data preprocessing, feature extraction, vectorization, and similarity measurement to generate accurate recommendations. Designed for simplicity and efficiency, the system demonstrates how machine learning can deliver meaningful, context-aware suggestions for modern digital reading platforms.
Author's Name: J.R.Arun Kumar, Janvi Prajapat, Anjali Chaudhary, Vanshika, Lakshya
Abstract— The proliferation of traditional attendance methods has led to an overwhelming volume of manual records for administrators to manage. The manual tracking process is often time-consuming, prone to human error, and may result in the overlooking of accurate attendance records. To address this challenge, this project introduces a web-based, AI-powered Attendance Management system. This application leverages advanced Computer Vision techniques to automate the identification of individuals using facial recognition. By capturing images and comparing facial features, the system quantitatively records a person’s presence, highlights essential identity matches, and identifies unauthorized entries. The primary objective is to streamline the attendance workflow, providing a rapid, objective, and efficient tool for both organizations and individuals to assess attendance accurately, thereby saving time and improving management accuracy.
Author's Name:J.R.Arun Kumar, Ms Tanu Saini, Tanisha Gupta, Unnati, Tamanna,
Abstract—This project presents a personalized movie recommendation system built using the MERN (MongoDB, Express.js, React, Node.js) stack integrated with a machine learning model. The system leverages content based filtering to suggest movies tailored to individual user preferences, such as previously liked genres, cast, and storylines. It utilizes movie metadata—including title, genres, description, actors, and directors—to generate feature vectors that drive similarity-based recommendations.
Author's Name: R.Anusuya, Mayank Saini, Chetna Soni, Love Kumar Saini, Kush Kumar Saini,
Abstract— This project develops a next-word prediction application for English typing using natural language processing techniques. The model is trained on a Wikipedia dataset of approximately 150,000 words to support diverse vocabulary across domains. Preprocessing steps include tokenization, stop word removal, and encoding, while feature extraction uses TF-IDF and tokenization. An LSTM-based model is implemented to capture context and predict the next word accurately. The system can also perform basic arithmetic operations by detecting numerical input and switching functionality accordingly. The trained model is deployed in a mobile application using React and TensorFlow Lite for real-time performance and improved user typing experience.
Author's Name:Mohit Sharma, Pankaj Yogi, Shailesh Paliwal, Nilesh Sahu, Yash Siddha,
Abstract— The rapid growth of digital documents has increased the demand for intelligent systems capable of retrieving accurate information from unstructured data. Traditional searchbased systems extract keyword-matching results, often lacking contextual understanding and semantic relevance. To address this challenge, this project proposes an Intelligent PDF Question Answering System leveraging Retrieval-Augmented Generation (RAG) integrated with Large Language Models (LLMs) to provide precise, context-aware answers from PDF documents. The system processes uploaded PDFs through an end-to-end pipeline consisting of document parsing, text chunking, embedding generation, vector storage, and retrieval-based response generation. Extracted text is converted into vector embeddings using transformer-based encoders and stored in a vector database such as FAISS or Pinecone. When a query is provided, the retriever fetches semantically relevant chunks, which are then passed to a generative LLM (e.g., GPT, LLaMA, or Mistral) to produce coherent, human-like answers grounded in the retrieved context. This hybrid approach reduces hallucinations and improves factual accuracy compared to pure generative models. The system features a user-friendly interface enabling document uploads, conversational query interaction, and citation-based output to highlight the exact source text.
Author's Name:Mohit Sharma, Nitin Khandelwal, Pradeep Kumar, Mohit Jajoriya, Deepanshu Jajoria
Abstract— This project presents an advanced Emotion-Based Music Generation system that leverages Machine Learning and Deep Learning to create personalized musical compositions aligned with human emotional states. Unlike traditional recommendation systems that only suggest preexisting tracks, this system dynamically generates original music by mapping emotional cues to musical attributes. The platform integrates multi-modal emotion recognition—utilizing facial expression analysis, speech sentiment extraction, and text-based emotion inference—to create a robust understanding of the user’s affective state. Facial images are processed using Convolutional Neural Networks (CNNs), whereas vocal audio signals are analyzed through MFCC feature extraction combined with Recurrent Neural Networks (RNNs) or Bi-LSTM layers for temporal modeling. Textual input undergoes natural language processing using Transformer-based architectures such as BERT to derive contextual emotional representations. Based on the predicted emotion category—such as happiness, sadness, calmness, anger, fear, or excitement—the system triggers the music generation module, which uses deep generative models
Author's Name: N Mahaviraswamy ,Chandana V , Jyothsna I , Kavana P , Kavya B J
Abstract—Abstract— Cloud Computing is the new field that was invented and developed during a period not so long ago. It has a lot of benefit such as decreasing the cost that the user needs, process the operation faster and keep the information secure. The security of cloud is the most important issue for several sensitive occupations, for web base information using this kind of computing does not need more than a computer and high speed internet to use application which is developed by cloud computing. This paper presents a model of web based information system at Amazon Web Services using visual studio to build a website with cloud data service which takes data from the instance in EC2 on Amazon Web Services (AWS), then lets AWS host data services from the cloud. The environment used, is EC2 (Elastic Cloud Computing) on AWS (Amazon Web Services) as host for an application which has several services to allow each user use the application separately and securely. AWS configuration and the way of using virtual server which its given by EC2 will be pointed out in this paper..
Author's Name: Mohit Sharma, Yash Yadav, Arpit Mukheeja, Vipul Meena, Naresh Kumar
Abstract— The rapid growth of digital media and social networking platforms has led to an increased spread of manipulated and fake images across the internet. With the advancement of image editing tools and AI-based generation techniques such as deepfakes, it has become extremely difficult to distinguish between real and fake images. These manipulated images can spread misinformation, create security threats, and negatively impact public trust in digital content. Traditional methods for detecting fake images are often manual, time-consuming, and ineffective against advanced manipulation techniques .This project presents a Fake Image Detection System using Deep Learning, designed to automatically identify whether an image is real or fake with high accuracy. The system utilizes deep learning models, particularly Convolutional Neural Networks (CNNs), to analyze complex image features such as textures, pixel distributions, and hidden patterns. These models are capable of learning subtle differences between authentic and manipulated images that are not easily detectable by human observation.The proposed system processes input images through multiple stages including preprocessing, feature extraction, model training, and classification. A labeled dataset containing real and fake images is used to train the model, enabling it to generalize and perform well on unseen data. The system provides quick and reliable predictions, reducing the need for manual verification.Overall, the project offers an efficient, scalable, and practical solution for detecting fake images. It demonstrates the potential of deep learning in digital image forensics and contributes toward improving the authenticity and reliability of visual content in modern digital environments.
Author's Name: Manoj Kumar Saini, Ritik Sharma, Mohit Kumar Saini, Nitin Saini, Sachin Poswal
Abstract— Fruit production is vital for global agricultural economies and food security, yet crop yields are frequently compromised by plant diseases. Traditional disease identification relies on manual visual inspection, which is labor-intensive, subjective, and prone to error, often delaying critical interventions. This paper presents an automated, highly efficient webbased diagnostic system for early fruit disease detection using Deep Learning. Specifically, the proposed model leverages Convolutional Neural Networks (CNNs) to accurately extract features and classify diseases from images of commercially significant fruits, including mangoes, bananas, and pomegranates. Going beyond standard image classification, the system uniquely integrates a recommendation engine that suggests targeted organic and Ayurvedic remedies for the diagnosed conditions, thereby promoting sustainable and ecofriendly farming practices. Deployed through a user-friendly web interface, this tool empowers farmers and agricultural researchers with instant, cost-effective decision support. By automating the diagnostic process and offering organic treatment plans, the proposed system ultimately aims to reduce reliance on harmful chemical pesticides, minimize crop loss, and enhance overall agricultural management.
Author's Name: Pradeep Kumar, Dhananjay Saini, Aditya, Hardik Jain, Kunal Yadav
Abstract— The rapid growth of digital image sharing has increased the need for secure and efficient image protection techniques. Traditional encryption methods are not well-suited for images due to their large size, high redundancy, and strong pixel correlation. To overcome these limitations, modern approaches using Deep Learning and Chaotic Maps provide enhanced security and performance .This project presents an Image Encryption and Decryption System using Deep Learning and a Chaotic Map, designed to ensure secure transmission of image data. The system uses chaotic maps to generate highly sensitive and random sequences for pixel permutation and diffusion, while deep learning techniques improve encryption strength and assist in accurate decryption .The system enables users to encrypt images into secure formats and decrypt them back to their original form using appropriate keys. It integrates key components such as image preprocessing, chaotic sequence generation, neural network-based transformation, and encryptiondecryption modules.The proposed solution improves data security, resists common cryptographic attacks, and provides an efficient framework for secure image communication in various applications.
Author's Name: Manoj Kumar Saini, Saurabh Saini. Rohil Khan, Pankaj Sharma, Sachin Jain,
Abstract— The Real Estate Price Prediction System is a Machine Learning–based solution designed to improve the accuracy and reliability of property valuation. Traditional methods often rely on human judgment and limited data, leading to inconsistent results. This project addresses these issues by using data-driven techniques to estimate property prices more objectively.The system is built using historical real estate data, including features such as location, area, number of bedrooms, and property age.The application is deployed using FastAPI for the backend, React with TypeScript for the frontend, and Supabase PostgreSQL as the database, ensuring scalability and real-time performance.Additionally, the integration of SHAP-based explainability helps users understand the impact of different features on price predictions. Overall, the system enhances transparency, reduces human bias, and provides a scalable, efficient, and user-friendly solution for accurate real estate pricing.
Author's Name: R.Anusuya, Priyansh Khandelwal, Udit Aggarwal, Sahil Khan ,Taniya Jangid.
Abstract— Communication barriers between hearing-impaired individuals and the general population remain a significant social challenge. This project presents an AI-powered Real-Time Sign Language to Text and Speech Conversion system designed to bridge this gap using computer vision and deep learning techniques. The proposed system captures static American Sign Language (ASL) gestures through a standard webcam without relying on external sensorbased hardware. Captured frames are preprocessed through Region of Interest (ROI) extraction, grayscale conversion, Gaussian blurring, adaptive thresholding, resizing, and normalization. A Convolutional Neural Network (CNN) trained on a structured ASL alphabet dataset classifies gestures with high accuracy, achieving 95–98% recognition under controlled conditions.
Author's Name: Arvind Sharma, Hemali Jain, Ishani Agarwal
Abstract— Video-based object recognition is an advanced technology in the field of computer vision and artificial intelligence that enables automatic detection, classification, and tracking of objects in video streams. Unlike traditional image-based systems, this approach processes continuous video frames, allowing the system to understand motion, behaviour, and temporal changes over time. It is widely used in applications such as surveillance systems, traffic monitoring, autonomous vehicles, healthcare monitoring, and smart city solutions. One of the major goals of this project is to overcome common challenges such as varying lighting conditions, occlusion, background noise, and real-time processing limitations. To address these issues, preprocessing techniques, feature extraction methods, and optimization strategies are applied to improve system performance. The system is designed to balance accuracy and speed, making it suitable for real-world applications. Furthermore, the system emphasizes scalability and adaptability, allowing it to be integrated with cloud platforms, edge devices, and IoT-based systems. It provides output in the form of annotated video frames with bounding boxes, labels, and confidence scores. Overall, this project demonstrates the potential of intelligent video analysis in automating tasks, improving decision-making To address these issues, preprocessing techniques, feature extraction methods, and optimization strategies are applied to improve system performance. The system is designed to balance accuracy and speed, making it suitable for real-world applications.
Author's Name: R.Anusuya, Chirankshi Sharma, Muskan Gupta, Vartika Chandrul, Subhiksha Jain
Abstract - The Vegetable Plant Disease System is an innovative automated solution designed to simplify and modernize the process of identifying and managing crop diseases in the agricultural sector. Traditional methods of disease identification — such as manual visual inspection by farmers or relying on scarce agricultural experts — are often time- consuming, prone to human error, and susceptible to delayed diagnosis, leading to significant crop yield losses. To overcome these limitations, this project utilizes advanced computer vision and machine learning technology, which offers a non-invasive, efficient, and highly reliable approach to detecting plant diseases automatically. The primary objective of this project is to develop a reliable, accessible, and user-friendly crop monitoring solution that saves time and reduces the economic impact of plant diseases.
Author's Name:Pradeep Kumar, Mohd.Yusuf Khan, Jatin Singh Chauhan , Gaurav Saini, Kuldeep Mudgal
Abstract— The modern recruitment landscape is characterized by a rapid increase in job postings and applicants, making efficient matching a significant challenge. Traditional keyword-based systems often fail to capture semantic meaning and user intent, leading to irrelevant recommendations and inefficiencies. To address this, the proposed Job Recommendation System (JRS) leverages advanced Machine Learning techniques to deliver personalized and context-aware job suggestions. It employs a hybrid approach combining Content-Based Filtering and Collaborative Filtering. The content-based component uses NLP techniques such as TF-IDF and cosine similarity to analyze relationships between user profiles and job descriptions, ensuring accurate matching based on skills and experience. Meanwhile, collaborative filtering utilizes user interaction data to identify patterns and recommend jobs preferred by similar users. The system also integrates a User-Interest Decay Mechanism to prioritize recent activities and a Missing Skills Module to suggest upskilling opportunities. Developed using Python, Scikit-learn, and Flask, the system demonstrates improved accuracy, scalability, and user satisfaction, offering an intelligent solution for modern recruitment challenges.
Author's Name: J.R.Arun Kumar, Sharafat Ali, Sahil Khan, Arman Khan,
Abstract— In today’s digital era, businesses receive a large number of customer enquiries through emails, chat systems, and online forms, making manual handling time-consuming and inefficient. This project, “Business Enquiries Using NLP and Deep Learning,” aims to automate the understanding and classification of customer queries using advanced AI techniques. The system analyzes textual enquiries to identify intent, category, and priority level by applying NLP techniques such as text preprocessing, tokenization, stop-word removal, and vectorization to convert raw text into meaningful numerical data. Deep Learning models like LSTM, RNN, and Transformer-based architectures are used for accurate classification and intent detection. The proposed system improves response time, enhances customer satisfaction, and reduces manual workload by automatically routing enquiries to appropriate departments such as Sales, Support, or Billing. It also incorporates word embedding techniques like Word2Vec, GloVe, and BERT to capture semantic meaning and context, enabling better understanding of user intent. Additionally, the system handles multilingual and noisy data, ensuring robust, scalable, and efficient performance in real-world business environments.
Author's Name: Arvind Sharma, Atul Pratap Singh, Chandrashekhar Sharma, Mohit Yadav, Ajay Kumar Saini
Abstract— The rapid growth of software development has led to an overwhelming amount of code being produced, making manual code review a time-consuming and error-prone process. Traditional review methods may overlook critical bugs, security vulnerabilities, and inefficiencies, affecting overall software quality. To address this challenge, this project introduces a web-based, AI-powered Code Reviewer system. This application leverages advanced Artificial Intelligence techniques to automatically analyze and evaluate source code for quality, performance, and adherence to coding standards. By examining code structure and logic, the system provides quantitative assessments, highlights potential errors, and identifies areas for improvement. The primary objective is to streamline the development workflow by offering a fast, consistent, and efficient tool for developers to enhance code quality, reduce debugging time, and ensure better reliability, thereby improving productivity and overall software performance.
Author's Name: J.R.Arun Kumar, Pratham Gupta, Kapil Yadav, Rakesh Kumar, Chetana Meena
Abstract—The rapid growth of Artificial Intelligence (AI) and Machine Learning has opened new frontiers in digital education. This project introduces an AI Teaching Assistant powered by a Retrieval-Augmented Generation (RAG) pipeline that uses video lectures as its primary knowledge base—a significant departure from conventional PDF-based RAG systems. The system transcribes video content using OpenAI Whisper, producing timestamped text chunks embedded into a local vector database. When a student poses a question, the system retrieves the most semantically relevant video segments and constructs a context-rich prompt for a Large Language Model (LLM) to generate accurate, grounded responses. The pipeline covers the full workflow: video-to-audio conversion, speech-to-text transcription with multilingual translation, metadata-aware chunking, embedding generation, vector similarity search, and LLM inference. Persistence is achieved through Joblib-serialized dataframes, eliminating re-processing on every run. The result is an intelligent tutoring system that answers student queries with direct references to specific moments in lecture videos
Author's Name: N Mahaviraswamy ,Chandana V , Jyothsna I , Kavana P , Kavya B J
Abstract—Abstract— Cloud Computing is the new field that was invented and developed during a period not so long ago. It has a lot of benefit such as decreasing the cost that the user needs, process the operation faster and keep the information secure. The security of cloud is the most important issue for several sensitive occupations, for web base information using this kind of computing does not need more than a computer and high speed internet to use application which is developed by cloud computing. This paper presents a model of web based information system at Amazon Web Services using visual studio to build a website with cloud data service which takes data from the instance in EC2 on Amazon Web Services (AWS), then lets AWS host data services from the cloud. The environment used, is EC2 (Elastic Cloud Computing) on AWS (Amazon Web Services) as host for an application which has several services to allow each user use the application separately and securely. AWS configuration and the way of using virtual server which its given by EC2 will be pointed out in this paper..
Author's Name: Pradeep Kumar, Shilendra Singh, Ms. Riya Kumari, Uttam Gupta.
Abstract— The aim of this project is to develop an all-in-one personal health companion, named Carebot, that integrates symptom analysis, medication management, and mental health tracking into a unified conversational interface. With the increasing complexity of personal health data, individuals often struggle with medication non-adherence and fragmented medical records. Therefore, early and accurate tracking of health symptoms and medication schedules is essential to prevent medical complications. This project employs an OpenAI-powered Large Language Model (LLM) integrated with a Next.js 15 framework to provide real-time health insights. The system automatically extracts medical entities from unstructured user chat to accurately log symptoms, mood, and medications. The proposed model specifically focuses on user-centric preventative care and ensures data accuracy through structured AI analysis. This system can be effectively utilized in home health monitoring, chronic condition management, and mental health support, providing a reliable tool to maintain health history and strengthen the patient-doctor relationship through data-driven insights.
Author's Name: Arvind Sharma, Sanjay, Laveesh Kumar Saini, Rohit, Yesh
Abstract— The rapid spread of misleading information on digital platforms has created an urgent need for automated fake news detection systems. This study presents a framework that leverages Natural Language Processing (NLP) and Machine Learning (ML) techniques to identify and classify fake news. Text preprocessing, feature extraction, are employed to capture linguistic patterns indicative of misinformation. Multiple supervised ML algorithms, including Logistic Regression, Random Forest, and Support Vector Machines, are evaluated to determine classification accuracy. Experimental results demonstrate that combining NLPbased features with robust ML models significantly improves detection performance, providing an effective tool to mitigate the societal impact of fake news.
Author's Name: Chandana V , Jyothsna
Abstract— Cloud Computing is the new field that was invented and developed during a period not so long ago. It has a lot of benefit such as decreasing the cost that the user needs, process the operation faster and keep the information secure. The security of cloud is the most important issue for several sensitive occupations, for web base information using this kind of computing does not need more than a computer and high speed internet to use application which is developed by cloud computing. This paper presents a model of web based information system at Amazon Web Services using visual studio to build a website with cloud data service which takes data from the instance in EC2 on Amazon Web Services (AWS), then lets AWS host data services from the cloud. The environment used, is EC2 (Elastic Cloud Computing) on AWS (Amazon Web Services) as host for an application which has several services to allow each user use the application separately and securely. AWS configuration and the way of using virtual server which its given by EC2 will be pointed out in this paper..
Author's Name: J.R.Arun Kumar,Aman Chauhan,Hitesh Sharma, Lokesh Kumar,
Abstract— The Email Spam Classifier project aims to develop a machine learning-based system capable of classifying emails into spam and non-spam (ham) categories. The system uses Natural Language Processing (NLP) techniques to analyze the content of email messages and extract meaningful features from the text data. Various machine learning algorithms such as Naïve Bayes, Support Vector Machine (SVM), and Logistic Regression can be applied to train the model using a dataset of labeled email messages.
Author's Name: J.R.Arun Kumar,Hemant Sharma,Dhruv Kumar,Abhishek Kumar,Hitesh Kumar
Abstract—This study presents an AI and machine learning–based approach for early detection of skin cancer using dermoscopic images. Convolutional Neural Networks (CNNs) are employed to automatically extract features and classify lesions as benign or malignant. The model is trained on labeled datasets, improving accuracy through data augmentation and preprocessing techniques. Performance is evaluated using metrics such as accuracy, sensitivity, and specificity. The proposed system aims to assist dermatologists by providing fast, reliable predictions and reducing diagnostic errors. Early detection through this method can significantly improve patient outcomes and enable accessible, cost-effective screening in remote or underserved areas.
Author's Name: Dr.Neeraj jain, Ms.Payal, Salim hussain, Parvej aalam,Kuljeet
Abstract - Emotion Detection Systems aim to automatically identify human emotional states using computational techniques, enabling machines to understand and respond to human feelings more effectively. This project focuses on designing and developing an intelligent emotion detection framework that analyzes facial expressions to recognize emotions such as happiness, sadness, anger, fear, surprise, and neutrality. The system employs image processing, feature extraction, and deep learning-based classification methods to detect facial landmarks, extract relevant expression features, and classify emotions in real time. By integrating techniques such as Convolutional Neural Networks (CNN) and Haar Cascade for face detection, the proposed system achieves high accuracy and robustness across varying lighting conditions and facial variations. This model can be applied in various domains, including human–computer interaction, mental health monitoring, education systems, security surveillance, and customer service enhancement. Overall, the emotion detection system demonstrates the potential of AI-driven technologies to improve communication between humans and machines. .
