Bot training and evaluation platform

Designed the bot-training experience for an enterprise AI automation platform used in regulated, document-heavy domains (finance, banking, insurance). Training previously lived in a separate internal tool; the MVP brought configuration, execution, and evaluation into a single platform. I owned the design of a guided, multi-step training workflow — from document selection and configuration, through execution with clear progress and failure states, to results and analytics supporting run-to-run comparisons. The core challenge was balancing ML-driven automation with usability and trust, enabling operators to iterate confidently while producing structured, reviewable outputs before deployment.