October 2026 Summaries

Kepler provides enterprise intelligence for investment firms and other high-stakes industries by combining code-based extraction and source citations for financial figures with language models that interpret questions and generate answers. To keep its public filings and customer documents current and traceable, the company uses Prefect to orchestrate continuous ingestion, preprocessing, and batch workflows across sources including SEC EDGAR filings, SharePoint, PDFs, spreadsheets, presentations, and HTML. Kepler replaced a homegrown orchestration framework and Temporal-based recovery setup after recurring issues with stuck runs, failed schedules, and maintenance demands, selecting Prefect for its APIs, documentation, hosting options, and support for dynamic tasks and parameterized runs. The migration took four weeks, with initial pipelines deployed within 10 days, and Kepler says pipeline changes now take one to two days rather than one to two weeks. It operates 32 production deployments across six pipelines, processes roughly 170,000 flow runs weekly and more than 20 million documents overall, and uses its pipelines to synchronize customer data sources, detect file changes, and prepare information for reliable cross-source queries.