Alfa Bank has developed an AI-based architectural consultant to address challenges in managing technical knowledge. The organization previously relied on fragmented data sources, including Confluence, system registries, and Architecture Decision Records, which hindered information retrieval efficiency.
The bank found that traditional Retrieval-Augmented Generation (RAG) methods were insufficient for complex architectural queries. These queries require more than simple text search, as the internal landscape involves numerous systems, modules, components, and integration layers.
To solve these issues, the team implemented a knowledge graph to structure data relationships. They also established a request routing system to navigate between various internal data sources, aiming to streamline onboarding and reduce the repetitive workload for senior architects.
