About Me

Hi there 👋 I'm EL MEHDI Hicham, an Artificial Intelligence Engineer passionate about Computer Vision, Deep Learning, and the application of AI in sports performance and analytics.

During my recent projects, I built complete video analysis systems for sports — combining object detection (YOLOv8/11), multi-player tracking (DeepSORT / ReID), jersey number OCR, and top-down projection for tactical insights.

I also explore OCR modeling (CRNN, Vision Transformers), event detection, and synthetic data generation to improve model robustness in real-world scenarios.

💼 Technical Skills

My expertise across AI, Computer Vision, and Software Engineering

Programming & AI

Python (PyTorch, TensorFlow, Keras)

95%

C/C++, Java

75%

FastAPI, REST API

85%

ML/DL (CNN, RNN, LSTM, Transformers)

90%

Computer Vision & Robotics

YOLOv8/11, Faster R-CNN, DeepSORT

90%

OCR (PaddleOCR, CRNN), Pose Estimation

85%

OpenCV, Homography, Camera Calibration

80%

ROS2, TensorRT, Real-time Inference, ONNX

75%

Generative AI & NLP

LLMS, RAG pipelines, Hugging Face

85%

LangChain, Prompt Engineering

80%

SQL-to-NL, Embeddings & Vector DBs

70%

Cloud & Infrastructure

AWS (S3, EC2, Lambda)

60%

Docker, Linux, Shell Scripting

80%

MLOps Basics (Deployment, APIs, Monitoring)

70%

Networking & Databases

TCP/IP

70%

SQL (MySQL, PostgreSQL)

80%

NoSQL (MongoDB), Vector DBs (FAISS, Pinecone)

75%

CDN

65%

Leadership & Collaboration

Team Supervision, Project Planning, Mentoring

85%

Git, GitHub, CI/CD Basics, Agile/Scrum

90%

Languages

Arabic (Native)

100%

French (Professional)

85%

English (Intermediate)

70%

📚 Experience & Education

R&D Engineer Intern - Computer Vision & AI

April 2025 — Present

SPORTSCORE - Football Analytics (Paris, France - Remote)

  • Designed and deployed a fully automated video analysis pipeline for futsal matches (object detection, multi-player tracking, homography, OCR, pose estimation)
  • Built an automatic frame classification module (YOLOv11m-cls) achieving 99.95% accuracy
  • Implemented complex action detection: pass classification (progressive, assist, under pressure), shots, dribbles, interceptions, goals
  • Optimized the pipeline, reducing inference time by 20% and improving ball detection robustness
  • Supervised and trained 4 interns, managing tasks, reviewing deliverables, and ensuring technical consistency
  • Delivered automated performance reports (heatmaps, pass maps, structured JSON exports, advanced stats)

Tech stack: PyTorch, YOLOv8/11, DeepSORT, OpenCV, OCR, Pandas, Python

Engineering Intern - End-to-End Event Spotting

October 2024 — January 2025

Université Ibn Tofail - Laboratoire SETIME (Kenitra, Morocco)

  • Developed a full pipeline: frame extraction, JSON annotations, temporal normalization, data augmentation (jittering, mixup), and train/val/test split (70/15/15)
  • Implemented RegNet-Y with Gate Shift Modules (GSM) for feature extraction and a Bi-GRU for global temporal reasoning
  • Experimented with multiple configurations (clip_len 16/50) using RegNet-Y and ResNet-50 on a 40GB dataset (7 matches from the 2024 Futsal World Cup)
  • Achieved a recall of 0.93 on the DRIVE class, with RegNet-Y outperforming ResNet-50

Tech stack: Python, PyTorch, RegNet-Y, ResNet-50, Bi-GRU, OpenCV, GPU (RTX 4060)

Internship - Content Delivery Network (CDN)

September 2023 — November 2023

Orange Business (Sale Al Jadida, Morocco)

  • Developed a Shell script to generate secure URLs with token-based authentication
  • Resolved a URL formatting issue and enhanced security by studying CORS headers, XSS, and CSRF attacks

Tech stack: Shell Scripting, JSON Web Tokens, wget

Master's Degree in Artificial Intelligence

November 2023 — September 2025

Ibn Tofail University (Kenitra, Morocco)

Specialized in:

  • Machine Learning & Deep Learning
  • Reinforcement Learning
  • Natural Language Processing (NLP)
  • Computer Vision
  • Generative AI

Key Projects:

  • Autonomous Vehicle with ROS2 & TD3 RL
  • OCR for Tifinagh Characters (Amazigh Heritage)
  • Diabetic Retinopathy Detection

Bachelor's Degree in Physics

2020 — 2023

Ibn Tofail University (Kenitra, Morocco)

General physics studies with a focus on mathematical modeling and computational methods.

Academic Project - Autonomous Vehicle

February 2025 — July 2025

ROS2 & Reinforcement Learning

  • Developed a ROS2–Gazebo simulation pipeline with URDF/xacro models
  • Integrated camera & LiDAR for obstacle detection and traffic sign recognition
  • Implemented a TD3 reinforcement learning agent for autonomous navigation
  • Combined trajectory planning, real-time control, and safe decision-making
  • Validated the system in diverse scenarios with reliable path following and collision avoidance

Tech stack: ROS2, Gazebo, Python, PyTorch, RViz, URDF/xacro, TD3 RL, Computer Vision

Academic Project - OCR for Tifinagh Characters

October 2024 — February 2025

Digitization of Amazigh Writings

  • Developed an innovative OCR system for Tifinagh handwritten script
  • Goal: preserving and digitizing the Amazigh cultural heritage using AI
  • Designed and implemented a hybrid CRNN-based model to improve character recognition accuracy
  • Role: Paddle OCR Developer and CNN Model Developer

Tech stack: Hybrid CRNN model, PaddleOCR, TensorFlow, PyTorch, OpenCV

Academic Project - Diabetic Retinopathy Detection

October 2024 — January 2025

Medical AI Application

  • Developed a Deep Learning model to detect and classify diabetic retinopathy from retinal images
  • Goal: facilitate early diagnosis to prevent vision loss and provide diagnostic aid for ophthalmologists
  • Preprocessed the APTOS 2019 dataset and implemented data augmentation techniques

Tech stack: EfficientNetB0, CNN, TensorFlow/Keras, Adam optimizer

🎯 Career Objective

My goal is to design and deploy intelligent vision-based systems that combine AI research, video analytics, and sports performance optimization.

I am currently seeking a Computer Vision / AI Engineer opportunity, where I can contribute to innovative, real-world applications that push the boundaries of what's possible with AI.

🚀 View My Projects