Architecting Conversational Intelligence for Enterprise Fintech

When I joined the team, Jeanie was already established as a retrieval system, but it was primarily in surfacing static FAQ content. While it handled basic tasks well, the next frontier was helping navigate the messy reality of natural language and ambiguous, multi-intent queries.

As part of this evolution, I worked alongside a cross-functional team to help pivot Jeanie from a content-delivery tool into a context-aware, LLM-powered assistant. My focus was on architecting the conversational logic and safety guardrails needed to handle complex transactions like balance inquiries and dispute initiations. By refining the intent taxonomy and the dialogue state machine, we are transforming Jeanie into a more intuitive, brand-aligned presence that could handle financial tasks within regulatory boundaries.