video1.0<iframe src="https://www.loom.com/embed/db24befd8f5b426d859e3fad80348a90" frameborder="0" width="1662" height="1246" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12461662Loomhttps://www.loom.com12461662https://cdn.loom.com/sessions/thumbnails/db24befd8f5b426d859e3fad80348a90-b36fde6a71eda5f5.gif444.16This Loom demonstrates LexAI, an application that improves marketing tagline translation while preserving meaning across languages. The presenter shows an input workflow that can translate or patch paraphrase using a German example, then lets users mark outputs as good or bad and provide feedback to reprompt the model; feedback is stored in a database for retraining. It also tracks daily DeepL token usage via an OpenAI layer that contextualizes results, originally using logins per user to isolate data. When user signups declined, the app switched to treating each browser session as a user, enabling privacy and admin visibility into active sessions and tokens used.