Mehdi Fathallah

Built a real-time computer vision system for lab tube and cap detection & matching using YOLOv8s-seg, achieving 82% mAP50–95 and 99% recall. Integrated PaddleOCR and OpenCV with < 1s latency (96% text accuracy), enabling automated QA in production. Optimized inference using OpenVINO and containerized deployment (Docker) for scalable CPU performance.

Developed an automated market intelligence system for banking data analysis, providing strategic insights for financial decision-making. Automated NLP and web scraping pipelines via Django REST Framework, APScheduler, and PostgreSQL. Integrated a RAG-based report generation module using LLMs for dynamic business insights.

Implemented an automated ETL to structure financial statements in Excel, reducing manual data processing time by 75%. Automated 500+ weekly financial reports via API-driven data pipelines.

Developed an automated invoice field extraction system based on CRNN (CNN + BiLSTM + CTC Loss), improving recognition accuracy by 30%.

Analyzed production data to identify critical operational patterns, reducing downtime by ~15 hours per week. Built real-time Tableau dashboards for KPI monitoring, improving operational visibility for production teams.