I'm an AI & ML engineering student based in Morocco, passionate about building systems that go beyond model training — from intelligent agent pipelines to end-to-end NLP workflows.
My work sits at the intersection of LLM engineering, multi-agent orchestration, and applied machine learning. I care about the full stack: clean architecture, solid theoretical grounding, and results that hold up in practice.
- 🔭 Currently: building a multi-agent RAG system that converts PDF specs into interactive HTML prototypes
- 📚 Deepening: agent memory, vector search strategies, and LLM evaluation frameworks
- 🎯 Goal: engineer AI systems that are not just functional, but reliable and explainable
Multi-Agent RAG System — PDF → Web Prototype
An autonomous pipeline that reads PDF specification documents (cahiers des charges) and generates fully functional interactive HTML prototypes. The system uses a
CRAgentto extract structured requirements, aCoderAgentto generate HTML, and anExecutorAgentto validate the output. It also includes an automated LLM evaluation pipeline that benchmarks multiple generation strategies to select the best-performing outputs — no manual review needed.
|
Autonomous LangGraph pipeline that parses PDF specs and generates interactive HTML prototypes via specialized agents. Includes an automated LLM evaluation system to benchmark and select the best generations.
|
Multi-class classification on the US Economic News dataset. Compared BoW vs TF-IDF with Logistic Regression and ANN. Applied SMOTE for class imbalance and corrected data leakage in the preprocessing pipeline.
|
|
Binary classification on the Wisconsin dataset using Logistic Regression, SVM, and Neural Networks. 97%+ accuracy with thorough evaluation including ROC curves, confusion matrix, and feature importance analysis.
|
End-to-end regression pipeline for retail sales forecasting. Feature engineering, model selection (Random Forest, XGBoost), and performance benchmarking with business-oriented metrics.
|
| Domain | Skills |
|---|---|
| LLM & Agents | RAG pipelines · Agent orchestration · Prompt engineering · Vector databases · LLM evaluation |
| Machine Learning | Supervised learning · Model evaluation · Feature engineering · Imbalanced data (SMOTE) |
| NLP | Text classification · BoW/TF-IDF · Preprocessing pipelines · Sequence models |
| Deep Learning | ANN · CNNs · Backpropagation · Regularization · Transfer learning |
| MLOps & Dev | Git workflows · Reproducible notebooks · Clean architecture · Docker |
📦 Agent memory & state management with LangGraph (StateGraph, conditional edges)
🔍 Vector search optimization — embedding strategies & chunking for RAG
📐 RAG evaluation metrics — faithfulness, relevancy, context recall
🐳 Containerizing ML pipelines with Docker
"The best way to understand a system is to build one."