EVO Bank’s virtual assistant needed a performance and scalability overhaul. The goal was to improve intent recognition, response quality, and cross-channel performance while reducing maintenance effort and enabling faster rollout of new features.

Responsabilities

• NLU analysis & optimization: Audited the assistant’s structure and performance to identify areas with the highest improvement potential.
• Model simplification: Refactored intents, entities, and training data to streamline the assistant’s logic and boost understanding.
• Conversational quality: Applied best practices to improve response accuracy and reduce fallback rates.
• Linguistic coordination: Led a team of linguists to ensure tone, clarity, and consistency across all channels.
• Impact tracking: Implemented metrics to measure improvements in both NLU accuracy and user experience.