Mohammed Haitham

Built AI-driven automation workflows and web-based AI tools integrating LLM components and OAuth-based API connections for production systems. Shipped internal and client-facing AI products end-to-end from prompt design through deployment.

Designed and trained an Attention U-Net + EfficientNet-B3 multi-class fetal abdominal-structure segmentation model and implemented an end-to-end maternal clinical pipeline. Built clinical guideline scraping, validated structured extraction, deterministic rule logic, Grad-CAM explainability, and provenance-traced report generation in SQLite.

Contributed to production AI services across Exavo’s product line, including transcription, email auto-reply drafting, social media automation, custom AI agents, and Shopify product automation. Built n8n automation workflows for LinkedIn publishing using Google/LinkedIn/OpenAI APIs and developed reusable agents and prompt flows in Microsoft Copilot Studio.

Contributed to ML/LLM prototypes and model evaluation in Python and built data preprocessing pipelines. Wrote clean, peer-reviewed code using core ML libraries.

Built LLM/RAG proof-of-concepts and retrieval pipelines integrating vector stores with conversational agents. Worked on prompt and agent design for multi-turn workflows.

Trained CNN-based models for product image recognition and retrieval and supported a visual-similarity recommendation system. Contributed to model development for image-based recommendations.

Grade: Very Good

Earned a BSc in Artificial Intelligence from Galala University (2021–2025). Graduation project: AI-based ionospheric forecasting using an LSTM pipeline for regional VTEC prediction with a real-time GUI for space-weather applications.