AI & Software Engineer
I build agentic AI assistants, retrieval systems and ML pipelines, and I measure them: every project below comes with the accuracy it reached and how I tested it.
Directors ask questions, apply filters, switch pages or request new visuals by text or voice. An 11-node LangGraph workflow routes each request, and Azure OpenAI answers are grounded in the active visuals, filters, measures and DAX. Entra ID and MSAL enforce row-level security.
A document assistant that reads PDF, DOCX, PPTX and XLSX files, and also charts and images through a local vision model. Hybrid search (dense plus BM25) with reranking gives answers cited to the page.
End-to-end pipeline on a year of minute-level sensor data. Feature engineering lifted the model from 60% to 92%, and XGBoost reached the final score. Shipped in Docker with CI/CD, Prometheus and Grafana monitoring, and email alerts.
SmolLM2 fine-tuned with QLoRA on 2M+ recipes, wrapped in a governance layer that checks calories, macros and allergens and logs every decision. Runs as Kubernetes microservices with Kafka and Jenkins CI/CD.
Designed and built the Power BI agentic assistant described above, from the Angular front end to the LangGraph backend.
Built and deployed the anomaly detection pipeline, with experiment tracking and production monitoring.
Built a full-stack asset and maintenance platform (MERN) with a Random Forest model that prioritizes and routes maintenance tickets.
Agentic AI system that helps households cut their energy use.
AI-driven tool to improve team productivity.
Engineering degree in Advanced Technologies, ENSTAB (2023 – 2026). Preparatory cycle in Mathematics and Physics, IPEIN (2021 – 2023).
Languages: Arabic (native), French (fluent), English (B2, TOEIC).
Open to AI engineering roles and freelance projects in agents, RAG and MLOps.