Skip to main content

My First Snowday: Notebooks, ML, and AI

By november 3, 2023maart 5th, 2024No Comments

Snowflake hopes to ride the AI wave

AI is booming, and Snowflake doesn’t want to get left behind. Their pitch: Snowflake’s core values and strengths have always revolved around data security, data governance, and data quality. What if you could apply cutting-edge AIs to your organization’s secure, high quality, well-documented, data? I think this is pretty exciting.

Basically, if an organization is curious about AI, Snowflake wants that organization to explore a unified data warehousing and AI vision through their data platform. The 2023 Snowday showcased an avalanche of new tools in the machine learning and artificial intelligence realms, as well as Snowflake’s vision of how all their latest products fit together. Let’s dive in!

Snowflake's AI Bento Box
Snowflake’s AI Bento Box

Snowflake Cortex: Simplifying Data Analytics

One of the most interesting additions, Snowflake Cortex is a set of serverless functions that aim to accelerate everyday analytics and AI app development. The service offers specialized ML and LLM models pre-tuned for specific tasks, such as sentiment detection, text summarization, translation, forecasting, and classification. (Most of these models are in public preview). In line with their business model, Snowflake wants to eliminate the complexities associated with managing infrastructure. Snowflake will worry about that — you only have to pay the bill.

Snowflake Cortex is also what is going to power an additional suite of bespoke AI tools (still in private preview). These include Snowflake Copilot (an LLM-powered coding assistant able to generate and refine SQL queries when prompted with natural language), Universal Search (an LLM-powered search-engine for quickly finding database objects), and Document AI (an LLM-powered data extraction tool for processing any documents, such as pdfs).

Snowpark Container Services: Packaging your AI apps

The modern development approach is to ship code in containers, ensuring portability and consistency across environments. Snowpark Container Services (in private preview) will allow Snowflake developers to ship their apps in containers. These containers are configurable with various hardware options and support files in every coding language. Snowflake wants to allow you to reliably ship AI-powered apps.

A diagram of  Snowflake Container Services
A diagram of Snowflake Container Services

Snowpark Notebooks: Experimenting with AI

As someone with a background in data science, I had to include this one. Snowpark notebooks are a cell-based programming environment, much like jupyter notebooks, that support Python and SQL code. When training AI models, or doing any kind of data science, notebooks are a fantastic development environment. They have their downsides — like any tool — but they are amazing for communication, documentation, experimentation, exploration and visualization. I think it’s very smart that they included these in their presentation on AI.

Final Thoughts

I came, I saw a llama, I conquered. I don’t think I’m the first to say that this year’s Snowday — my first Snowday — was absolutely packed with features. As someone who learned of Snowflake as primarily a data warehousing solution, I’m impressed with the strength of their pitch for AI integration, and excited to see these features deployed!



Leave a Reply