Case Study: a GCC-based security preparedness company achieves transparent AI document search with ScienceSoft

a GCC-based security preparedness company validates AI document search with ScienceSoft using 4 LLM-as-a-judge metrics

ScienceSoft developed a proof of concept for an AI document search assistant for a GCC-based security preparedness company. The client needed a more reliable and transparent way for its internal users to search its large knowledge base of security-sensitive documents, as conventional keyword search was insufficient for their complex queries.

The solution was a RAG-based assistant that utilized a hybrid search pipeline combining BM25 and vector search, powered by the Llama 2 model. This approach supported both precise keyword matching and semantic search, with answers grounded and linked to their source documents. ScienceSoft delivered a validated technical foundation for an enterprise rollout, giving users a faster way to retrieve and verify information from their large document knowledge base.