Crafting Intelligent Solutions

About Hamza Boughanim

AI and Software Engineer with deep expertise in Large Language Models, Computer Vision, and full-stack development. 3+ years building intelligent, production-ready solutions.

Education

• Master's in AI & Virtual Reality — Ibn Tofail University (2023–2025)
• Bachelor's in Computer Science — Ibn Tofail University (2019–2023)

Frequently asked questions

Who is Hamza Boughanim?

Hamza Boughanim is an AI & Backend Engineer based in Morocco with 3+ years of experience building production LLM, RAG, computer vision and MLOps systems. He holds a Master’s degree in Artificial Intelligence & Virtual Reality from Ibn Tofail University and currently works on AI document automation for the insurance sector.

Where in Morocco is Hamza Boughanim based?

He is based in Kenitra and works between Kenitra and Rabat, in the Rabat-Salé-Kénitra region of Morocco (North Africa). He works remotely with clients across Morocco, the Maghreb, Europe and worldwide, in the Africa/Casablanca time zone (UTC+1).

Is Hamza Boughanim available to hire as an AI engineer?

Yes. He is open to remote, hybrid and international AI/ML engineering roles as well as freelance and contract projects. The fastest way to reach him is hamzaboughanim06@gmail.com or the contact form at hamzaboughanim.com/contact.

What AI systems has he shipped to production?

A document-AI platform for a major Moroccan insurance group combining hybrid OCR at 91.3% accuracy, local Phi-3 classification and NER at over 90% precision, and pgvector semantic search across 760+ documents; a satellite computer vision pipeline reaching mAP50 0.836 with YOLO11m and mIoU 0.804 with a fine-tuned SAM1 ViT-H; a Zero-Trust LLM inference backend; and a RAG chatbot built on FastAPI, ChromaDB and LangChain.

What makes an AI engineer in Morocco or North Africa different?

Trilingual document work is the clearest difference. Administrative, legal and insurance corpora across Morocco and the Maghreb mix Arabic and French on the same page, where English-tuned OCR and NER pipelines degrade badly. Engineering specifically around that produced the 91.3% hybrid-OCR result. Working with tight GPU and privacy budgets — local Ollama inference, quantised models, embedding caching — is also default practice rather than an afterthought.

Which AI and machine learning technologies does he work with?

PyTorch, TensorFlow and scikit-learn for modelling; LangChain, Ollama, the OpenAI API, AWS Bedrock and Google Gemini for LLM integration; pgvector, ChromaDB and Pinecone for vector search; YOLO11, SAM, SAHI and OpenCV for vision; Tesseract, EasyOCR and PaddleOCR for OCR; and FastAPI, Docker, Kubernetes, DVC and GitHub Actions for serving and MLOps.

Does he work with self-hosted or local LLMs?

Yes. He runs Phi-3, Llama-2 and Mistral locally through Ollama, chosen deliberately for privacy-sensitive and GDPR-constrained deployments where data cannot leave the perimeter, alongside hosted APIs where those are appropriate.

What languages does he work in?

Arabic (native), French (professional) and English (professional) — which matters technically for document AI and OCR across North African corpora, not only for communication.