Sho Tanaka

The Model Context Protocol (MCP) is emerging as a standard for connecting AI agents with external tools. While LLM-based tools are flexible, their non-deterministic behavior makes it difficult to meet requirements for stability, repeatability, and auditability in production systems. In such cases,...

Unlock insights trapped in unstructured data using Cortex AI Functions. In this hands-on lab, you'll build batch extraction pipelines that process documents, images and audio — all with SQL. Load files to Snowflake stages, then use AI_PARSE_DOCUMENT to extract text from PDFs, AI_EXTRACT to pull structured fields from text and images, AI_CLASSIFY to categorize content, and AI_TRANSCRIBE to convert audio to text. Build automated pipelines with Dynamic Tables that process new data as it arrives. By the end, you'll transform unstructured files into queryable tables — extracting entities, classifying documents, and structuring data for analytics or downstream AI.

Moving ML from experimentation to production shouldn't require rebuilding from scratch or navigating lengthy development cycles. In this hands-on session, you'll build an end-to-end fraud detection model using an agentic approach in Snowflake. Starting with natural language prompts in Snowflake CoCo from Snowflake Notebooks, you'll train a model, register it to the Snowflake Model Registry and deploy it for real-time inference, all without leaving Snowflake. Walk away with practical techniques for creating production-ready prediction endpoints and scaling online inference workloads for fraud detection and beyond.