London Snowflake User Group (LSUG)

We're in for a treat! The first of two events in september, to pick back up after a summer break.
To start September off, we have 2 talks from Snowflake Engineering, which is a treat since they are traveling across Europe to meet LSUG in London.

## Max Heimel | Taking a Peek Under the Hood of Snowflake's Query Performance Features

One of Snowflake's core architectural advantages is its ability to deliver high-performance analytics without sacrificing elasticity. Central to this is our separation of compute and storage: data lives in cloud storage, accessible by any compute node on demand, while transactional consistency is maintained through a centralized metadata service. This separation enables near-instant spin-up of compute resources, but it also means that reading data comes at a cost. To keep queries fast and efficient, we need to be smart about what data we bring onto the node and what we skip. That's where Snowflake's metadata system comes in: each micro-partition is richly annotated with statistics and summaries that let our engine quickly prune away irrelevant data before scanning even begins. In this talk, we'll take you on a deep dive into how Snowflake organizes data internally, how we use metadata to make intelligent pruning decisions, and how you can take advantage of features like clustering and search optimization to speed up your own workloads.

## Marco Slot | Postgres vs. Unistore: When to use which

Snowflake recently announced the acquisition of Crunchy Data and the Snowflake Postgres managed service. With that, Snowflake now offers two operational database products: Unistore and Postgres. This talk will go into architectural differences between Postgres and Unistore, and discuss the application characteristics that determine whether you should use one or the other. We'll also look at example scenarios for hybrid- and operational workloads and how they map into Unistore and Postgres.

## Paul Gray, Preventx | From Ground to Cloud – How Prevent X Transformed Data Sharing and Reporting with Snowflake

Paul shares how Prevent X partnered with Slalom to define a tailored data strategy, built around real user needs across lab, finance, and product teams. He’ll walk through how they consolidated data into Snowflake, with privacy and governance built in from day one.
From commercial insights to sentiment analysis using Cortex, Paul will show how teams across Prevent X now use data to make faster, smarter decisions—without needing technical skills. He’ll also touch on culture change, self-service, and what’s next on their data journey.

## Keith Belanger | From Chaos to AI-Ready: How DataOps Makes Snowflake Work for AI

AI is your organizations hot priority, yet Gartmer projects over 60% of AI projects will fail. Why? Because the data and the ecosystem aren't ready, and legacy delivery methods can’t keep up.
In this session, we’ll explore how DataOps transforms Snowflake from a powerful platform into a reliable engine for AI success. You’ll learn why half-baked, stitched-together data pipelines are no match for the speed, control, and governance that today’s AI initiatives demand, and how a modern DataOps practice/solution can close the gap. Whether you’re just getting started with AI or scaling across the enterprise, this session will reframe what it truly means to be “AI-ready” and why modern DataOps is no longer optional.

## Colm Moynihan | The Top 10 Snowflake New Features that get Colm out of bed