Aditya Gohil

– Designed & trained a custom YOLOv8 model via PyTorch transfer learning for aerial human & vehicle detection, achieving 92% mAP@0.5. – Engineered a low-latency OpenCV + TensorRT inference pipeline for real-time edge deployment on UAVs (<40 ms/frame). – Built an end-to-end MLOps data pipeline (ingestion, augmentation, stratified splitting) to improve model robustness across aerial conditions.

-Designed and maintained internal MySQL databases with normalized schemas and optimized queries, serving as the data backbone for research team tooling and reporting workflows. -Built ETL automation using Node.js scripts, replacing manual ingestion processes and enabling reliable, scheduled data pipelines — directly supporting engineering team productivity.