Mohamed Laouar

At DataLab ComGroupe Building and deploying an LLM-powered assistant at scale : ▷ Led backend architecture for an LLM-powered assistant using microservices communicating though REST APIs (FastAPI) ▷ Designed and built a scalable RAG system handling complex, heterogeneous documents ▷ Built an automated RAG evaluation pipeline using classification-based metrics to measure chunking, retrieval, and augmentation quality and compared it to state of the art accessible solutions ▷ Enhanced the RAG system with memory, reranking  and hybrid retrieval to increase precision beyond semantic search alone ▷ Integrated agentic decision-making — orchestrating agents for intent recognition, business-specific logic, and response formatting ▷ Ensured production-grade code quality, scalability, and reliability - supporting 2k+ users on constrained infrastructure Deploying GenAI in a large corporate environment like BNP is challenging: the technology is new, and the software team behind it has to dig deep into the existing IT ecosystem to make it work and get it validated by management. Under these circumstances our team was one of the first three to get a GenAI product live and available to users.