Sudhir Gajre

Finance back office teams spend significant time on manual, document-heavy workflows like remittance reconciliation and proof of delivery validation. We present how we deployed two production-grade automation use cases for Lippert, saving 250+ hours monthly with over 90% model accuracy. We first cover remittance reconciliation from lockbox PDFs spanning hundreds of pages. The solution uses a deterministic extraction pipeline that segments documents, normalizes headers, cleans OCR noise, validates line items and applies customer enrichment to produce finance-ready tables and exception-focused audit trails. Next, we detail proof of delivery automation, where standard OCR struggles with crossed-out quantities and signatures. This use case combines computer vision and LLMs using Azure Document Intelligence and OpenCV smart cropping with a safety-first validation layer. We share design tradeoffs, failure modes and lessons learned from deploying finance agents in production.

Taxonomy generation is a challenge across industries such as retail, manufacturing and e-commerce. Incomplete or inconsistent taxonomies can lead to fragmented data insights, missed monetization opportunities and stalled revenue growth. In this session, we will explore a modern approach to solving this problem by leveraging Databricks platform to build a scalable compound AI architecture for automated taxonomy generation. The first half of the session will walk you through the business significance and implications of taxonomy, followed by a technical deep dive in building an architecture for taxonomy implementation on the Databricks platform using a compound AI architecture. We will walk attendees through the anatomy of taxonomy generation, showcasing an innovative solution that combines multimodal and text-based LLMs, internal data sources and external API calls. This ensemble approach ensures more accurate, comprehensive and adaptable taxonomies that align with business needs.