Computer vision system that detects and extracts text from vehicle license plates using OpenCV edge detection and Tesseract OCR. Processes vehicle images to identify plate regions, highlight contours, and read alphanumeric characters automatically.
Deep learning model for automated medical X-ray analysis. Detects and classifies bone fractures using CNN architecture trained on medical imaging datasets to assist radiologists.
Real-time object detection system for road safety compliance monitoring. Detects whether vehicle occupants are wearing seat belts using YOLO architecture on live camera feeds.
Smart agriculture sensor system with automated alerts. Monitors soil moisture levels in real-time using IoT sensors and triggers automated irrigation alerts to optimize water usage.
Production-ready NLP chatbot automating high-frequency customer support queries. Reduces manual support workload by 60%+ through intelligent query classification and automated responses.
Production-ready Retrieval-Augmented Generation web app. Chat with your documents using a local LLM, FastAPI async backend, ChromaDB vector store, HuggingFace embeddings, Ollama llama3.2, streaming chat UI and full admin panel. Fully self-hosted with Docker support.
Full-stack HR and Continuing Professional Development platform for UK care-sector organisations. Manages staff profiles, mandatory training, PDF certificates, policy compliance, reference requests, and VLE credentials, with role-based access for administrators and staff.
End-to-end n8n automation workflow for Inspire London College custom LMS. Automates student account creation, course enrolment, payment instalment setup, assignment deadline configuration, and branded credential email delivery, all driven by Google Sheets data.
Suite of n8n workflows automating student registration, course enrolment, WhatsApp notifications, Google Sheets sync, and email triggers for UK education clients. 80%+ manual work reduced.
8-class deep learning classifier for gastrointestinal diseases from endoscopy images. Built with EfficientNetV2S transfer learning on 8,040 real endoscopy images, achieving 95% test accuracy. Final Year Project at University of Haripur.
End-to-end ML pipeline for HR data, fuzzy name deduplication, feature engineering, scikit-learn classification, exportable performance reports via FastAPI.
Full-stack web platform offering 40+ free browser-based tools, file converters, social media utilities, text tools, and developer tools. Built with custom PHP CMS, SEO-driven blog engine, and admin dashboard. Self-managed on a Linux VPS with Nginx and MariaDB.
Computer vision system that detects smoke and its origin point in images and video using YOLO object detection. Trained across 6 experiments on a custom Roboflow dataset, achieving mAP50 of 0.679 with YOLOv11n, supporting early fire warning and surveillance automation.