http://hdl.handle.net/123456789/132 Tue, 28 Apr 2026 08:26:23 GMT 2026-04-28T08:26:23Z http://hdl.handle.net/123456789/20585 Automated Credit Scoring 03-134221-031, Muhammad Wahaaj Tauqir Commercial clients (business owners, retailers) of wholesale distributors often rely on credit lines to purchase goods, especially during economic crises when banks hesitate to lend to retailers. This affects the sales of distributors, prompting them to offer products on credit. However, when distributors extend credit, they face the risk of customer default. They need a reliable method to evaluate the creditworthiness of their customers to minimize financial loss. To address these challenges, we propose an automated AI driven credit scoring system that automates the evaluation process by analysing several features including customer purchase behaviour, transaction history, and repayment patterns to the interpretable scorecard. The system integrates generative AI–powered variable classification, automated preprocessing, Auto-Monotonic Fine Binning, and advanced ensemble modeling and instant generation of points-based scorecard. Deploying the solution as a secure cloud-native application ensures rapid, accurate, and fully auditable credit decisions significantly improving risk management and operational efficiency for wholesale distributors. Dr. Muhammad Saqib Sohail Thu, 01 Jan 2026 00:00:00 GMT http://hdl.handle.net/123456789/20585 2026-01-01T00:00:00Z http://hdl.handle.net/123456789/20586 DineSmart: Al Powered Chatbot 03-134221-012, Haider Naeem DineSmart shifts traditional food ordering system to smart chatbot based ordering system. It combines intelligent ordering platform which understands customer intents, extract entities properly, manages complex orders and restaurant operations with robust backend built on powerful Flask-Python backend with flutter used for interactive and user-friendly frontend. This system also integrates multiple technologies including Google’s Gemini 2.5 Flash LLM for conversational AI, Stripe for dummy payment processing, MongoDB for data storage and retrieval and Google Maps API for precise food delivery. Our platform provides dual purpose architecture: providing customers with new ordering experience with giving restaurant administration real time order management, inventory control and business analytics. By doing so DineSmart sets new market standards in the field on food industry. Dr Muhammad Saqib Sohail Thu, 01 Jan 2026 00:00:00 GMT http://hdl.handle.net/123456789/20586 2026-01-01T00:00:00Z http://hdl.handle.net/123456789/20955 AI Assisted Caretaker For Neurodegenerative Disorder 03-134202-013 ASAD ALI, 03-134221-033 NAZAR ALI Neurological disorders encompass a broad spectrum of diseases affecting the brain, spinal cord, and nervous system. Almost 600 neurological disorders exists globally. Early diagnosis is critical for mitigating the societal and economic impact of NeuroDegenerative Disorders (NDD). Alzheimer’s Disease (AD) is one of the most prevelent neurodegenerative disorder. Additionaly, there is no cure to AD. The early and precise diagnosis of AD remains a significant challenge in clinical neuro-science. Neuroimaging is one of the significant biomarker for early diagnosis of AD. The research work compraises of Magnetic Resonance Imaging (MRI) data acquired from Kaggle. Dataset consists of 4 classes: Mild Demented (8960), VeryMildDemented (8960), Moderate Demented (6464), NonDemended (9600) , 33706 samples in total. Pillow (PIL) is used for data preprocessing to greyscale, resize and normalize the image samples. For classification Convolutional Neural Network (CNN) is used with 3 convolutional blocks and 1 hidden layer. Grid Search is used for hyper-parameter tuning. Accuracy achieved is 97%. RAG with HuggingFace Transformer model embeddings is used to generate context-aware and intelligent responses. Notification system to notify patients and doctors about important events. Appointment Scheduling module is used to schedule and keep track of appointments between doctor and patients. Prescription table is used in system to keep record of patient prescription after an appointment. Dr. Iram Noreen Thu, 01 Jan 2026 00:00:00 GMT http://hdl.handle.net/123456789/20955 2026-01-01T00:00:00Z http://hdl.handle.net/123456789/20956 Vogue AI: Next Gen Fashion Stylist 03-134221-043 UNAIZA MUKHDOOM, 03-134221-015 MEHAK The fashion and apparel industry faces complex challenges in delivering personalized styling service that account for the relationship among physical attribute, circumstances, and individual inclinations. The whole process of dressing up has numerous interdependent variables that affect what the individual wears. These are the corroboration of skin tone with weather, the fittingness of events, and the need of culture and the self-expression of aesthetics. The traditional form of shopping and the existing fashion technologies fail to provide comprehensive guidelines on these dimensions. They in particular do not focus on such important aspects as skin tone, undertone, face shape, and body proportions. The absence of this gap results in bad choices, ineffective wardrobes, and insecurity, as well as repeated fashion mistakes. Users do not know what colors, outfits and trends to use without the help of real time experts. Modern strategies such as using the Flutter framework for cross-development, ensuring compatibility across both android and iOS devices, marked the development process. Forthcoming features include and a modern Material 3 design theme to enhance user engagement and functionality. The thesis develops an intelligent AI-powered fashion recommendation system for Pakistani market. Mobile application uses OpenCV for facial features analysis including skin tone, undertone, eye color, hair color and face shape detection. It provides scientifically-backed seasonal color palettes and personalized styling with digital wardrobe system. Application scrapes fashion data from Pakistani brands Khadi, Sapphire, Gul Ahmed, Nishat Linen and Akaram. Flutter-Dart app with Firebase. Intelligent Dialog flow chatbot in English-Urdu. Fit Me analyzes body type and scores of the user pic. The report outlines the technical strategies adopted to develop these features, the methods used in addressing them and the results of the built application. It also discusses recommendations made regarding future updates and improvements, which focus on the improvement of accessibility, contents, and the overall user experience. Dr Junaid Nasir Qureshi ( Thu, 01 Jan 2026 00:00:00 GMT http://hdl.handle.net/123456789/20956 2026-01-01T00:00:00Z