Senior AI/ML Engineer with 5+ years of experience delivering production-grade AI systems across Computer Vision and Natural Language Processing. Currently completing an M.Sc. in Computer Science & AI (GPA 3.97/4.0) while building agentic systems, conversational AI, and speech intelligence solutions in industry. I have designed and deployed end-to-end pipelines spanning object detection, segmentation, player re-identification, medical imaging, OCR, RAG-based agents, ASR, and TTS — consistently bridging cutting-edge research with scalable, real-world impact.
Designing and developing advanced AI systems with a focus on Agentic architectures, conversational AI, and speech technologies. Responsibilities span the full development lifecycle — from research and prototyping to production deployment — across the following areas: Agentic Systems (multi-step reasoning pipelines and autonomous AI agents), Chatbots (context-aware, domain-specific conversational interfaces), Automatic Speech Recognition (ASR), Text-to-Speech synthesis (TTS), and Data Evaluation (building evaluation frameworks and quality benchmarks to measure model performance and ensure reliability at scale).
Worked on real-time projects involving object detection, pose estimation, segmentation, and developing an automatic annotation tool. Handled deployment phases, utilized EC2 instances and internal servers, and followed Agile methodologies with Microsoft Azure DevOps.
Responsible for handling OCR tasks, including data collection, annotation, object detection, text recognition, and deploying models, while also developing automation tools like workflow management systems to support the team. Additionally, utilized AWS services such as SageMaker and EC2, and employed Agile methodologies (Azure DevOps, Linear, Jira) to ensure smooth project execution.
Implemented research projects in data science, computer vision, deep learning, and machine learning, from data collection and cleaning to model development and deployment.
Developed software solutions for medical technology applications, focusing on reliable and efficient system implementations.
Built an AI engine to process football match videos and automatically extract key performance metrics, including physical measurements, pass types, attack patterns, possession, ball losses/recoveries, set pieces, and other football actions. The system utilized a wide range of Computer Vision techniques such as object detection, tiny object segmentation, camera calibration, action recognition, object tracking, and player re-identification. It was deployed using FastAPI with MongoDB integration for efficient data storage and retrieval.
Developed a voice-interactive chatbot that enables users to input queries via speech and receive spoken responses. The system was built as a multi-stage pipeline comprising Speech-to-Text (STT), a Retrieval-Augmented Generation (RAG) system powered by a Large Language Model (LLM), and Text-to-Speech (TTS). Custom components were created for embeddings, STT, TTS, and the RAG system, achieving excellent domain-specific results. FAISS and Qdrant vector databases were used to ensure efficient semantic search and retrieval. To support training, custom annotation tools were built to streamline dataset creation and labeling. The solution was deployed using both Flask and FastAPI, allowing for flexible integration across platforms.
Developed a custom OCR pipeline for Egyptian national ID cards using specialized text detection models and three dedicated recognition models tailored for ID numbers, manufacturer numbers, and general text. The system achieved high accuracy in text extraction and recognition, outperforming well-known OCR engines such as Qwen and Azure. To enhance security, liveness detection and fraud detection mechanisms were integrated. The solution was deployed using FastAPI, enabling smooth integration into existing systems.
Led a project to analyze a client's sales history and customer data, delivering key insights to support their marketing strategy and loyalty program. Using Power BI, I developed a dynamic, scalable dashboard that visualizes customer retention, segmentation, top-selling products, and ABC analysis for both customers and products, with monthly and quarterly metric tracking. Additionally, I designed feedback analysis forms to monitor customer complaints and contributed to refining the marketing approach based on ongoing results.
Developed a system that uses pose estimation to track body keypoints and calculate joint angles for precise evaluation of exercise performance. Custom logic was implemented for each exercise type (Squats, Push-ups, Pull-ups) to differentiate between successful and unsuccessful repetitions. When a rep is identified as incorrect, the system pinpoints the specific error and delivers corrective feedback through voice instructions. The setup includes both front and side cameras to capture a comprehensive range of angles, enabling accurate form analysis and correction.
Built a robust pipeline for the multiclass classification of Retinopathy of Prematurity (ROP) into five distinct categories. The process began with the collection of medical retinal images from hospitals, followed by annotation and data refinement using custom filtering tools. Grad-CAM analysis showed that the classifier focused on critical pathological regions, which led to the integration of segmentation for added context. A U-Net++ model was trained on a custom dataset to segment retinal blood vessels, while ridge-like features were segmented using the zero-shot MedSAM model after detection with YOLOv11. These segmented features were fused with the original images prior to classification, resulting in a significant boost in accuracy and highly reliable outcomes.
Accumulative grade: Excellent with Honors with a GPA 3.97/4.
Overall grade: Excellent with Honors with a GPA 3.99/4. Overall percentage: 97.44%. Ranked 1st in the department out of 120 students.
Comprehensive programming training and certification.