Multilingual Agentic RAG Framework
A modular Arabic/French framework for asking questions over structured and unstructured information using retrieval and AI agents.
Evaluation Metric
Achieved 92% Answer Relevancy and 100% Context Recall on the demonstrator's evaluation dataset.
The Problem
Building a useful assistant over heterogeneous information requires more than sending documents to an LLM. Data must be prepared, indexed, retrieved and provided with enough context for grounded responses. The framework also needed to handle French and Arabic.
Solution & Contribution
Developed a modular framework combining ETL pipelines, document vectorization and indexing, retrieval modules and AI agents for contextual response generation. Built an insurance-domain conversational assistant as a demonstrator to validate the framework.
My Contribution
Framework architecture and implementation across data preparation, vectorization / indexing, retrieval and agent-based response generation; insurance assistant demonstrator.
Technologies Used
Interested in collaborating?
Get in touch with Khalid to discuss applied AI engineering, LLMs, or agent systems.
kh.regui@gmail.com