AI application development is moving beyond experimentation. Developers are now expected to build intelligent applications and agents that are reliable, secure and ready for production. This keynote introduces how Snowflake helps developers turn enterprise data into AI-powered experiences. It explores how to modernise data foundations, make structured and unstructured data AI-ready, and build governed applications and agents using Snowflake Cortex, Snowpark, Streamlit and Snowflake Intelligence. Attendees will learn how to connect data sources, apply business context, integrate AI into existing workflows, and move from prototype to production while maintaining governance, performance and cost control.

Peek under the hood to learn how hybrid table architecture delivers mission-critical performance and reliability for your applications while reducing build time and complexity through full native Snowflake database integration. We also dive into recent optimizations that have delivered a 10x improvement in transactional workload performance.

Have multiple copies of the same data? Learn how Snowflake is delivering interoperability without compromise, making it easier to architect for multi-engine access on a single data copy, across more use cases, with governance policies that persist across engines at the table, row, and column level. In this What’s New session we’ll cover the latest features expanding what you can do with Iceberg v3, how Snowflake Storage for Apache Iceberg(tm) makes it easy to unlock interoperability without bucket management, new Horizon Catalog capabilities that deliver universal governance for multi-catalog environments and persist fine grained access policies across engines, and how to share and access open data within and outside of Snowflake’s boundaries.

You’ve invested in Iceberg. But managing catalogs, storage, compaction and security across engines still falls on your team. Snowflake meets you wherever you are and reduces your operational burden at every step, from instant connectivity to zero-ops. We’ll show you how it works, all fully open via IRC, with no disruption to what you already have.

Business Ontology is a governed model of business terms, metrics, entities, and their relationships in Horizon Catalog. It organizes concepts by business domain and links them to the semantic views, tables, and columns that represent them. Customers can create or import definitions through Cortex Code and browse them in Snowsight as a glossary or graph.

Explore how to build and deploy applications with Snowflake App Runtime across three use cases: interactive data exploration on governed data, workflow apps with shared state and approvals, and AI-powered apps and agents using Cortex Agents and MCP tools. Learn how to streamline app development through the describe, build, deploy and iterate lifecycle, work with the app.yml manifest, and manage in-place upgrades at a stable URL.

Make your data AI-ready with the powerful combination of dbt Projects and Snowflake Semantic Views. This session demonstrates how to leverage both SQL and Python DataFrame models in dbt for flexible data engineering that appeals to Spark developers and SQL practitioners alike. Explore semantic layers, the Open Semantic Initiative (OSI) and best practices for managing semantic view lifecycles in Snowflake. Watch Snowflake CoCo’s AI coding agent accelerate dbt Project development in real time, and learn to automate deployments with Snowflake CLI. Walk away with a complete, modern approach to building AI-ready data pipelines that empower both AI models and analytics tools.

Unstructured data (text, images, audio and documents) is the most valuable and least activated layer of the enterprise data stack. This session covers unstructured data preparation and search at enterprise scale, including: The end-to-end data enablement pipeline: How OpenFlow, AI enrichment and Cortex Search compose into a production workflow that takes raw unstructured data to searchable, agent-ready assets in minutes. Snowflake Search Stack: Four modalities covering interactive RAG, batch matching, exhaustive analytical queries over unstructured corpora and federated retrieval from a single agent, all built on Snowflake and scaling to billion-sized indexes.

This session covers how to build AI agents on Snowflake with Cortex Agents, and how to run them in production. Topics include:
– Building and operating AI agents: connecting an agent to structured data (Cortex Analyst) and documents (Cortex Search), writing agent instructions, and managing versions, access and performance.
– Building production-ready RAG applications: preparing and chunking documents, improving search results, and returning answers with source citations.
– AI observability and evaluation: monitoring agent activity, creating a test dataset, and measuring answer accuracy using Snowflake’s built-in evaluation tools.

Agentic AI can turn data into action, but only if it can find the right data and understand it. That’s a challenge since most enterprise data isn’t sufficiently self-describing. In this session, we’ll define what AI-ready data looks like and show how Horizon Catalog provides the critical context layer that helps AI know your data. Horizon Catalog delivers quality-aware context directly to AI agents — so they don’t just access data, they understand it. It also catalogs data from external systems, captures end-to-end lineage and usage patterns, and automates documentation and semantic modeling. Join us to see how a strong metadata foundation enables AI that’s more accurate and trustworthy.

Organizations are increasingly looking for cohesive AI-powered data products that combine structured, semi-structured and unstructured data. This session details how to build, package and securely share AI-powered multimodal data products. We define “interoperable AI-powered multimodal data products” and discuss architectural considerations for reliable product evolution with live demonstrations. Learn how to leverage catalog-linked databases with open table formats, multimodality features and AI layering (semantic views, Snowflake Cortex Search, managed MCP server and so forth) alongside industry-leading AI and data collaboration capabilities to deliver comprehensive data products at scale.

In this virtual hands-on lab, you will get to experience how Snowflake Online Feature Store provides low-latency, key-based feature retrieval for real-time ML inference. This hands-on lab demonstrates how to build an end-to-end machine learning workflow using the Online Feature Store.

The introduction of the Snowflake-managed MCP server has generated significant customer interest and practical deployment questions. This session directly addresses the most common challenges we’ve observed across numerous customer engagements, specifically focusing on server setup, secure authentication, tool integration and establishing external client connectivity. Attendees will gain practical guidance and validated best practices derived from real-world deployments to ensure a seamless and successful MCP implementation.

This session shows how builders can use an AI coding agent to migrate and modernise legacy data pipelines on Snowflake, and finish with data that is ready for AI. Topics include:
– AI-assisted migration: assessing legacy SQL, stored procedures and ETL code, converting them with SnowConvert, and using Cortex Code to fix the issues it couldn’t convert automatically and explain each change.
– Modernising, not just lifting and shifting: having the agent rebuild batch jobs as incremental Dynamic Tables on open Iceberg tables, loading data from multi-cloud sources (AWS S3 and Azure Blob), and validating results against the source.
– AI-ready and governed by design: adding data quality checks, sensitive-data classification and masking, and a semantic view so the modernised data can be used safely by BI and AI.
– Using agents safely in engineering: restricted session scopes, least-privilege roles, human review of every change, and auditing agent activity through query history.

Bring your code to your data, and go from ideas to apps in minutes. Snowflake Apps is evolving to bridge the gap between your code and the enterprise data ecosystem. We’ll unveil the launch of native support for modern frontend frameworks like React and showcase new integrations with AI-powered development workflows. Discover how to move from a natural language prompt to a secure, production-ready app in record time, and hear from companies that are already building faster on Snowflake.

Your enterprise data already holds the answers. The challenge is getting AI to understand it and act on it securely. In this session, our head of AI product shares Snowflake’s vision for the agentic enterprise: where every user has a personal work agent that builds context from your data automatically, works across structured and unstructured sources, and operates within full governance. See how customers build sales call analyzers, automate financial reporting and deploy customer-facing agents in minutes. Whether you’re building agents with Snowflake CoCo or using them through Snowflake CoWork, you’ll leave knowing how to put governed, model-agnostic AI to work.

What if you could go from a blank notebook to a monitored production model using just natural language? Snowflake CoCo brings agentic AI to the Snowflake ML lifecycle by training models, engineering features with Snowflake Feature Store, registering and versioning models in Model Registry, deploying real-time inference endpoints on Snowpark Container Services, running batch predictions at scale with ML jobs, tracking experiments and more. In this session, we demo the full journey: explore data, train a model, build feature pipelines, deploy to a GPU endpoint and set up monitoring — all driven by an AI agent through conversation. See how agentic ML development collapses weeks of MLOps into minutes.

AI-ready data isn’t just clean and governed. It has to be served the moment a user, or an agent acting for them, asks for it. This talk takes a query that already runs fine for analysts and turns it into a real-time, user-facing product feature with Snowflake Interactive Analytics. You’ll watch a standard warehouse and an interactive one handle the same query under the same load, then learn what interactive tables and warehouses do differently, and where they fit.

Operational data is valuable for analytics, but moving it from Postgres into an analytics platform often requires complex ETL pipelines, external tooling and ongoing maintenance. In this virtual hands-on lab, you will use Snowflake Postgres data mirroring to automatically replicate operational data into Snowflake for analytics — without any extra infrastructure. You will create a sample IoT application schema, set up data mirroring in minutes, validate initial and ongoing replication, explore the $live view for low-latency reads and review the $changes feed for row-level audit history.