Mohamed Elagamy

AI Engineer specialising in building and deploying production-grade AI systems — from real-time computer vision pipelines to end-to-end RAG platforms and Arabic NLP models. Delivered a fully containerised equipment-monitoring system for EagleVision, including YOLO fine-tuning, multi-object tracking, Kafka streaming, and a live Gradio dashboard. Built an Arabic Tashkeel Transformer from scratch achieving 3.5% DER on 70 EPUBs.

Comfortable across the full stack: model training in PyTorch, API deployment with FastAPI, and production infrastructure with Docker and Kafka.

AI Engineer at EagleVision Company

Technical assessment for EagleVision Company — production-grade, microservices-based CV system monitoring construction equipment in real time.

• Fine-tuned YOLOv12n on a custom construction dataset; selected nano over medium variant after A/B testing — same accuracy, significantly faster inference for real-time requirements
• Implemented DeepSORT for multi-object tracking, maintaining persistent Track IDs across video frames
• Designed region-based Optical Flow to detect articulated machine-part motion (e.g. excavator arm digging while tracks remain stationary), outperforming LSTM and 3D CNN alternatives in speed with zero labelled data required
• Built a state classifier (ACTIVE/INACTIVE) and activity classifier (Digging, Swinging, Loading, Dumping, Waiting) per tracked machine
• Streamed structured JSON telemetry via Apache Kafka → ingested into TimescaleDB; served per-machine utilisation stats via a real-time Gradio dashboard

B.Sc. in Computer Science and Artificial Intelligence – Fayoum University (2020-09 – 2024-06)