Rehman Ullah

AI Engineer Intern SID Lab UET 05/2024 – 09/2024 | Peshawar •Contributed to research-driven AI projects in academic settings •Preprocessed and annotated datasets for ML experiments •Trained and evaluated ML models using Python and deep learning libraries •Worked collaboratively on NLP tasks and transformer-based solutions AI Automation Engineer Blue Health Institute 03/2025 – 05/2025 | Lahore •Developed automation scripts in Python for healthcare data workflows •Assisted in integrating AI-based solutions into existing systems •Analyzed structured and unstructured data to support automation tasks Hate Speech Detection of Roman Urdu Using Transformer Methods Final Year Project •Developed a robust NLP system for detecting hate speech in Roman Urdu using transformer models like BERT •Addressed linguistic variability, data scarcity, and preprocessing challenges •Achieved high accuracy by fine-tuning models and optimizing the pipeline Breast Cancer Detection Using Machine Learning •Built a classification pipeline using Random Forest and SVM on medical imaging data •Emphasized early detection and precision to support clinical decision-making •Optimized features and reduced false positives through model tuning Misogyny Detection NLP Classification •Trained BERT, RoBERTa, and ML models (LR, SVM, RF, NB). •Applied advanced preprocessing for Roman Urdu toxicity detection. •Built a complete pipeline for evaluation and deployment. Roman Urdu Chatbot University Administration •Developed a chatbot for university admin queries as a learning project. •Used LSTM, BERT, RoBERTa for intent detection and text understanding. •Created preprocessing for Roman Urdu slang and mixed language input. Diabetes Prediction System •Trained 9 ML models (LR, RF, SVM, NB, KNN, GB, XGBoost, DT, Linear Reg). •Built 4 deep models: MLP, CNN, LSTM, and GRU. •Compared all models using accuracy, F1 score, and ROC-AUC. Strawberry Disease Detection •Trained a custom YOLOv8 model for leaf disease detection. •Applied edge AI to optimize model for real-time mobile/field use. •Used Explainable AI (XAI) visualizations (heatmaps) for model clarity. Car Classification •Trained a decision tree classifier for car type prediction. •Processed and labeled dataset for supervised learning. •Evaluated performance using accuracy and a confusion matrix.