What you'll learn

Data teams are being asked to use LLMs for classification, enrichment, summarization, triage, and operational automation, but production systems need more than a prompt connected to a pipeline. In this course, Designing Data Systems with LLMs, you’ll gain the ability to design LLM-powered data systems that fit cleanly into end-to-end data architectures. First, you’ll explore where LLMs belong in data pipelines and how to separate deterministic components from AI-driven components. Next, you’ll discover how to design prompting, RAG, and agentic workflows for finance and e-commerce style use cases. Finally, you’ll learn how to orchestrate, trigger, monitor, and validate LLM-enabled pipelines using tools such as Apache Airflow, LangGraph, FastAPI, DuckDB, PostgreSQL, and pgvector. When you’re finished with this course, you’ll have the skills and knowledge of LLM-powered data system design needed to integrate generative AI into production data architectures without losing reliability, data integrity, or operational control.