Generative AI is moving from experiments to production, and organizations need professionals who can build systems on real enterprise data — structured tables alongside unstructured documents, contracts, and transcripts.
Whether you're a student, early-career professional, or experienced practitioner, this four-course program takes you from Snowflake fundamentals to working AI agents. Created by Snowflake's developer advocates, it emphasizes hands-on learning with in-product exercises, so you finish with skills you can apply on the job — not just concepts.
You will learn:
How to work with Snowflake's core objects including virtual warehouses, databases, UDFs, stored procedures, and Snowpark DataFrames
How to run common AI tasks — summarization, translation, sentiment analysis, classification — with Cortex AI functions
How to select and prompt a foundation model, and fine-tune one to distill a larger model or match a desired style
How to build retrieval-augmented generation apps on unstructured documents with Cortex Search
How to build Text-to-SQL apps that answer natural-language questions from structured data with Cortex Analyst
How to give users a front end for those apps with Streamlit
How to design AI agents that plan, select tools, and synthesize insights across multiple data sources
How to evaluate agent reliability, write orchestration instructions, and set up observability
How to integrate agents with external applications using Model Context Protocol
Applied Learning Project
Every course is built around in-product exercises, and the work compounds across all four.
You'll start by creating Snowflake objects, writing UDFs, calling your first Cortex AI function, and editing a Streamlit app. From there you'll tackle common AI tasks and fine-tune a model to a desired style. Then you'll build a RAG app over unstructured documents with Cortex Search and a Text-to-SQL app over structured data with Cortex Analyst, each with a Streamlit front end.
You'll finish by building two AI agents. The first uses deal metrics and transcripts to configure semantic views, build a search service, and connect both to an agent that analyzes win rates, surfaces concerns, and synthesizes insights. The second handles insurance claims with multi-step workflows and external integration via Model Context Protocol.
All projects use realistic datasets mirroring enterprise scenarios. By the end you'll have working apps and agents you can show as proof of your skills.
