video1.0<iframe src="https://www.loom.com/embed/d0ef8320ee2742aea3a0685f7cf7a315" frameborder="0" width="1068" height="801" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>8011068Loomhttps://www.loom.com8011068https://cdn.loom.com/sessions/thumbnails/d0ef8320ee2742aea3a0685f7cf7a315-67250a97b24cdd4a.gif289.218This Loom walkthrough explains the Lead AI challenge pipeline for document classification and extraction. The project uses three stages: deterministic classification of incoming documents, LLM-based extraction of values into Type Records, and a separate calculation and validation step to match services to compatible rates or refuse when needed. The author runs the full pipeline in offline mode for reproducible results, with inference minimized to avoid extra LLM calls, and reports the instruction process takes less than one minute. Output is returned as JSON in structure output format, with example logs showing resolved and unresolved booking and packlist cases.