I am Ilyas Fardaoui, an AI Engineer and ENSAM Rabat student focused on building intelligent systems that solve real-world problems. I design and ship production-oriented AI workflows, including retrieval-augmented generation pipelines, multimodal search engines, and data-intensive NLP systems. My strongest edge is working on underrepresented language spaces, especially Darija and Arabic NLP, while bridging academic ideas with practical deployment constraints. I believe in turning research-heavy ideas into practical systems that teams can use, scale, and trust.
- 🔭 Working on: Agentic RAG pipelines with LangGraph and real-time multimodal retrieval systems
- 🌱 Learning: Advanced agent workflows, tool-use patterns, and LLM evaluation frameworks
- 🤝 Open to: AI engineering roles, research collaborations, and open-source contributions in Arabic/Darija NLP
| Project | Description | Stack | Domain | Status | Stars |
|---|---|---|---|---|---|
| VisualIndexer | Production-grade multimodal search engine combining CLIP embeddings, OCR extraction, and vector similarity retrieval for image-text discovery. | Python, CLIP, Tesseract, ChromaDB/FAISS, Streamlit | Multimodal Retrieval | Active |  |
| darija-dataset-builder | Scalable data pipeline for building high-quality Moroccan Darija corpora ready for fine-tuning and evaluation workflows. | Python, NLP, HuggingFace Datasets, Pandas, MinHash | Moroccan NLP | Active |  |
| Sepsis-Detection | Early-warning ML pipeline for sepsis risk prediction using ICU clinical features and interpretable boosting models. | Python, scikit-learn, XGBoost, Pandas | Healthcare AI | Active |  |
| Reconnaissance-Faciale-Eigenfaces | Classical computer vision implementation of PCA/Eigenfaces for robust face representation and recognition experiments. | Python, OpenCV, NumPy, PCA | Computer Vision | Active |  |
| GOLD-TRADING-AI | AI-assisted market analysis framework for gold forecasting, signal generation, and risk-aware strategy exploration. | Python, Time Series, ML, Visualization | FinTech AI | Active |  |
| YouTube-Sentiment-Analysis | End-to-end sentiment pipeline for YouTube comments including preprocessing, modeling, and actionable polarity insights. | Python, NLP, scikit-learn, Pandas | Social NLP | Active |  |
| Role | Organization | Period | Key Achievements |
|---|---|---|---|
| AI Engineering Intern | Ministry of Agriculture Morocco (MAPMDREF) | Internship | - Built a multimodal semantic search engine for practical retrieval use cases. - Developed OCR pipelines for document and image text extraction. - Improved vector-based retrieval workflows for higher relevance and speed. |
| Vice President | Fatal Error Club - ENSAM Rabat | Leadership Role | - Led the club's technical direction and execution strategy. - Mentored members across software and AI engineering tracks. - Co-organized CURSOR Meet-up with 190+ participants and technical sessions. |
- Open-source Darija LLM evaluation benchmark
- Agentic document search assistant (RAG + tool use)
- Arabic multimodal dataset for visual question answering
I'm always open to interesting collaborations, AI research projects, and engineering roles. Don't hesitate to reach out!