
Today, we're joined by Shreya Shankar, a PhD student at UC Berkeley to discuss DocETL, a declarative system for building and optimizing LLM-powered data processing pipelines for large-scale and complex document analysis tasks. We explore how DocETL's optimizer architecture works, the intricacies of building agentic systems for data processing, the current landscape of benchmarks for data…