Mujahid waheed

■ Engineered a real-time workplace safety monitoring system using YOLOv8 across multi-camera RTSP feeds, detecting PPE violations and
unsafe behaviours with sub-100 ms inference latency.
■ Owned the full ML pipeline — Roboflow annotation, class-balanced augmentation, model training, hyperparameter tuning, and ONNX
export — reducing inference overhead by ~40%.
■ Containerised all AI inference services with Docker, enabling reproducible, environment-agnostic deployments across staging and
production infrastructure.

■ Automated data collection pipelines via Selenium and BeautifulSoup, ingesting large-scale socioeconomic datasets and eliminating manual
research effort for recurring reporting cycles.
■ Built and deployed classification models (SVM, Decision Trees) with feature engineering and outlier detection; delivered interactive
Streamlit dashboards for policy-driven stakeholder insights.
■ Applied PyTorch and TensorFlow to NLP and computer vision research tasks, producing reproducible experiment results and model
evaluation reports for think tank publications.