AI/Data Science Lab

Runnable Snowflake AI and data science labs with sample data, step-by-step scripts, model workflows, evaluation checks, and cleanup sections.

Articles

• Build a Cortex Analyst Semantic Model Lab: A Cortex Analyst lab for creating governed sales data, drafting a semantic model, asking natural-language questions, validating SQL, and cleaning up.
• Build a RAG Workflow With Cortex Embeddings: A retrieval-augmented generation lab using Snowflake Cortex embeddings, vector similarity, answer generation, quality checks, and cleanup.
• Build a Snowflake Feature Engineering Lab for Churn Models: A runnable lab for preparing customer churn features in Snowflake with sample events, data quality checks, train/test splits, and cleanup.
• Build a Snowflake Feature Store Lab for Reusable ML Features: A Snowflake Feature Store lab for defining reusable customer features, creating training data, validating freshness, and cleaning up.
• Deploy Open Source Model Workflows With Snowpark Container Services: A deployment-oriented lab showing how to package an open-source model workflow, register model metadata, evaluate inference logs, and clean up.
• Extract Structured Fields From Documents With Cortex: A document-style Gen AI lab that extracts invoice fields from sample text, validates structured JSON, checks accuracy, and cleans up.
• Govern and Evaluate Gen AI Outputs in Snowflake: A lab for governing Gen AI prompts and outputs with role-aware inputs, masking, audit tables, quality checks, and cleanup.
• Monitor Model Drift and Retraining Signals in Snowflake: A model monitoring lab that compares training and production feature distributions, flags drift, tracks prediction quality, and cleans up.
• Run Time Series Forecasting Checks in Snowflake: A forecasting lab that prepares sample demand data, builds a simple baseline, evaluates forecast error, explains results, and cleans up.
• Run an End-to-End Snowflake Notebook Experiment: A Snowflake Notebook lab for organizing an ML experiment with setup, profiling, training, metrics, decision checks, and cleanup.
• Train and Register a Snowpark ML Classification Model: A Snowflake Notebook lab for training a churn classifier with Snowpark ML, evaluating predictions, registering the model, and cleaning up the demo.
• Use Cortex LLM Functions for Support Triage: A runnable Cortex AI lab that classifies support tickets, summarizes customer issues, evaluates output quality, and cleans up the demo.

What belongs here

• runnable Snowflake data science and Gen AI labs
• sample data that can be created directly from SQL worksheets or Snowflake Notebooks
• model training, evaluation, inference, registry, and lifecycle examples
• Cortex AI, LLM, embedding, vector search, and RAG workflows
• secure, governed AI patterns that are practical for enterprise Snowflake environments