Muawiya Shakeel

Architected and deployed an end-to-end Visual AI Surveillance Platform for people counting, speed-violation detection, facial recognition, vehicle classification, and ANPR across 40 live RTSP cameras. Built concurrent DeepStream inference pipelines and a React/FastAPI dashboard with Oracle event logging, containerized deployment, and CI/CD.

Built a production CNIC OCR and scanner application using Google Vision API with Urdu-to-English translation and automatic CNIC formatting. Packaged and deployed the solution across KTM and KGM to reduce manual entry and improve reporting speed.

Supported supervisor-led AI/ML research by implementing experiments, preparing datasets, and building reproducible development prototypes. Developed and optimized models for anomaly detection, predictive analysis, and classification, and explored GPT/LLaMA with RAG for document retrieval and chatbot development.

Developed Python automation scripts for production data analysis, improving workflow efficiency with data science, machine learning, deep learning, and NLP use cases. Integrated SQL-based inventory tracking and contributed to Flask/FastAPI REST APIs and an operational analytics dashboard.

Pursuing an MS in AI/AI-ML research with a focus on federated learning and explainable medical AI. Authored a manuscript on dataset-size-weighted GradCAM aggregation for full-stack AI applications and achieved an AUC-ROC of 0.8235 on NIH ChestXray14.

Worked as a research assistant supporting supervisor-led AI/ML projects through dataset preparation, experiments, and reproducible prototypes. Developed anomaly detection and predictive/classification models (reaching 95%+ in selected experiments) and explored GPT/LLaMA and RAG workflows.

Completed a BS in Artificial Intelligence (2020–2024). Built and deployed a final-year malware detection system (“The Malware Hunter”) using XGBoost/Scikit-learn and a FastAPI/Flask/Streamlit interface, achieving 90% classification accuracy.