Snowflake Developers

Every team running AI at scale needs to answer: 🎯 How do you know the output is correct? ⚡ How do you make AI efficient enough to run at scale? Over the past year, our Snowflake AI Research team partnered with Brown University, UCLA, University of Chicago, and UC Santa Barbara on 5 papers to answer both and push the state of the art for Cortex AI Functions: 1. Perfect classification accuracy at 3.2x lower cost 2. Tests analytical correctness, exposing cascading pipeline errors 3. Maximizes semantic ranking accuracy within any compute budget 4. Up to 19x reduction in AI query overhead 5. 95%+ accuracy by routing easy rows to a smaller model 👉Read the full breakdown: https://bit.ly/4fQPzMF

How do you manage LLM safety once your AI app hits production? Snowflake Data Superhero Abhishek Mittal shows how Cortex Guard helps enforce it, filtering outputs inside your Snowflake environment using Llama Guard 3, all with a simple SQL or Python config. A practical look at how to move from ad hoc safeguards to more reliable AI safety in production: https://bit.ly/4fbp18I

Big update for developers building enterprise AI applications. Claude Opus 4.8 is now available in public preview across Cortex Code, Cortex Agents, and Cortex AI workflows in Snowflake, bringing stronger reasoning and more capable agentic execution directly into governed development environments. Details: https://bit.ly/4ebC2hq

This guide breaks down how graph queries work in Postgres using Apache AGE, and why that approach can make more sense for connected data, from fraud detection to recommendation systems. Explore how graph queries fit into modern data workflows: https://bit.ly/4a9fPy7

Every company has invoices, contracts, and receipts sitting in cloud storage. The data inside is critical, but extracting it means stitching together OCR, writing brittle parsers, and maintaining it all when layouts change. 📄 CoCo takes you from raw documents to a structured, queryable table in a single prompt. ✅ Parses PDFs from any vendor layout using AI Extract ✅ Pulls vendor names, invoice numbers, line items, and totals automatically ✅ Loads everything into a Snowflake table ready to query The document processing pipeline you used to build by hand now comes from one prompt. 👇 Watch James Cha-Earley walk through it in under 3 minutes. Get the link to full demo in the comments.

How fast can you move from raw code to production? In the latest Codestrap CxC episode, Snowflake's Head of Developer Experiences Umesh Unnikrishnan breaks down how Cortex Code is helping builders move from idea to production faster with a data-native AI coding agent built for real enterprise workflows. From developer productivity to the rise of the agentic enterprise, this conversation dives into what modern DevEx should actually look like. 🎬 Watch the full episode: https://lnkd.in/gMfc5KHf

AI costs and access can get out of control quickly. Between multiple models, agents, and teams experimenting at once, it’s easy to lose track of who can use what, how much it costs, and how to keep things governed as usage scales. Snowflake Data Superhero Pierre-Mickaël Chancrin breaks down why governance is becoming essential for working with Cortex Code, and how to think about access control, cost management, and scaling AI responsibly. Check it out: https://bit.ly/3RCzXlW

As Iceberg adoption grows, one question keeps coming up: Which catalog should own your tables? That choice is not easy to undo. The catalog becomes the source of truth for metadata, access control, and how every engine reads and writes your data. Check out this break down on how to think about that decision, including open standards, governance at the catalog layer, bidirectional interoperability, and operational independence: https://bit.ly/4nKsaP2

If you’ve ever tried to scale heavy batch inference workloads over large amounts of unstructured data, you know the pain: managing distributed infrastructure and handling heavy data ingress/egress. We’re making that easier to run at scale in Snowflake with job-based batch inference on Snowpark Container Services (SPCS), powered by Ray. One API call, distributed execution on GPUs, and support for both structured and unstructured data, all without managing infrastructure. If you’re running large-scale inference workloads or migrating them into Snowflake, check this out: https://bit.ly/3PF4XRA

Want to simplify data engineering and build autonomous pipelines for AI agents? We've got the hands-on session for you. Join us on May 27 to learn how to use Cortex Code to turn prompts into pipelines, create a Dynamic Table for auto-managed orchestration, and power it all with an interactive Streamlit app directly in Snowflake: https://bit.ly/49cP8s6