Hire the Best Computer Vision Engineers
in Morocco

I'm a certified AWS Machine Learning Engineer with a strong foundation in AI, machine learning, and data science, having completed numerous successful projects ranging from complex workflows to real-time analysis systems. My technical expertise includes Python programming, deep learning frameworks, machine learning libraries, and cloud deployment platforms, allowing me to build efficient data-driven solutions from model development to cloud deployment. As a Kaggle Expert, I bring competitive problem-solving abilities to deliver impactful results across various domains including LLMs, computer vision, and data engineering

I build AI systems that solve hard, real-world problems. Recently I developed a dual-camera computer vision system that analyzes bodybuilding exercise form in real time, tracking body keypoints, scoring movement quality frame by frame, and delivering actionable feedback through a clean interface. The client came back with more work. That's the standard I hold myself to. My focus is on the intersection of AI and practical engineering: → Computer vision and pose estimation systems → Machine learning model development and deployment → LLM integration and AI-powered applications → Python backends and data pipelines → End-to-end delivery, from model to working product I have an engineering degree and a strong foundation in mathematics and algorithms. I don't just run pre-built models. I understand what's happening inside them, which means I can debug, optimize, and adapt when things get complicated. I work with clients who have problems that are a step beyond what a typical developer can handle. If that's you, let's talk.

Most OCR problems are not actually OCR problems. The document is rotated, curled, photographed at an angle, or scanned badly and the extraction fails before it even starts. That's the part I work on. I'm a computer vision engineer at a fintech in Morocco. My job is processing bank statements for automated credit analysis, which means I deal with the worst documents you can imagine: phone photos of paper statements, faxed scans, pages photographed half-folded on a desk. I built the system that straightens them out before extraction. Some concrete numbers from that work, because everyone on here claims "99% accuracy": Tested my rectification pipeline on 1,441 real bank statement images that I labeled by hand. Median angle error 0.28 degrees, 85% of pages corrected to within half a degree. Trained deep learning models (U-Net, SegFormer, and others) on 90,000 synthetic statement images I generated myself, because no public dataset exists for this. Best model: 3.4px endpoint error. Built the batch infrastructure to run this on thousands of documents without babysitting it. Re-runnable, doesn't corrupt files if it crashes mid-job. Stack is Python, OpenCV, PyTorch, Azure Document Intelligence for the extraction side. If you need bank statements, invoices, or receipts turned into Excel/CSV/JSON or you already have an OCR setup that keeps failing on bad scans send me a couple of sample files. I'll look at them and tell you what accuracy is realistic and how I'd do it. No charge for that, and if I don't think I can get you good results I'll say so.

I'm a PhD in computer vision. I can do both research and engineering of machine or computer vision systems. I have developped two machine vision systems for defect detection and and size measurement of products or components. I can leverage traditional image processing, machine learning, and advanced deep learning to create robust and efficient solutions. In terms of research, I have published various papers at top conferences and journals. I volonteer as a scientific reviewer and teaching computer vision on my YouTube channel. My latest experience is implementing a modern object tracker published in a top journal from scratch, then adapting it to 360° videos using advanced geomtrical projection. Please refer to my portfolio, YouTube, and Google Scholar account for more details.

I fix the 3 reasons why Computer Vision models fail in production: high latency, brittle lighting changes, and massive cloud hosting bills. Instead of deploying heavy, over-engineered academic frameworks that slow down your system, I build lightweight, high-velocity Python engines using OpenCV, PyTorch, and YOLO optimized specifically for real-world visual data. What I deliver: • Custom Object Detection & Real-Time Tracking • Zero-Lag Video Stream & Image Preprocessing • Production-Ready Deployments via Docker If you have a messy dataset, an unstable video stream, or a model that is eating up too much hardware compute, send me the details. I build vision systems that work under real constraints.