The Future of Snowflake: Data-Native Apps, LLMs, AI, and more
Written by
Ajit Monteiro, CTO & Co-Founder
Published
June 27, 2023
OneSix is excited to be attending the world’s largest data, apps, and AI conference: Snowflake Summit. The opening keynote had a lot of exciting announcements for the world of data, and continued strategy of rolling out AI and Data-Native App capabilities to their platform . Below are some of the things we found most interesting:
A more complete Data-Native Apps Stack with Container Services
Streamlit and Snowpark have been available for a while now. However, the addition of Snowpark Container Services helps us fully realize Snowflake’s Data Native Apps goals.
Continuing their vision of moving all your company’s data into Snowflake as a governed secure environment, you can now use it in a more cloud platform centric way. Snowpark Container Services allows you to run Docker containers which can then be called by Snowpark; you now have a UI solution (Streamlit), a data-native coding solution (Snowpark) and a way to run legacy applications (Snowpark Container Services) in the Snowflake cloud. You can then easily distribute and monetize these apps through their marketplace.
Use Case Example: A client of ours wanted to use Python OCR services that leverage Tesseract. In the past this was difficult to do since you cannot install Tesseract in Snowpark, Snowpark Container Services will allow us to install Tesseract in a container, and use a wrapper Python library like Pytesseract in Snowpark to leverage it.
Large Language Models and Document AI
It seems like everyone has been talking about large language models (LLMs) lately, and it’s not surprising that Snowflake had some big announcements around it. It was interesting to learn about Snowflake’s partnership with Nvidia to power their Container Services, as well as their first party LLM service.
They also released a feature called Document AI that allows you to train their large language model with your documents and then ask questions against them. This UI based approach allows you to modify the model’s answers to your questions about the document. Those modifications feed back into the LLM, training it to work better on your company’s data.
Streamlit becoming a more robust app UI platform
Streamlit has been historically marketed as a ML focused UI tool. However new features are making it a more viable platform for hosting general apps on Snowflake. A notable feature that have been released this year are editable data frames, including copying and pasting from Excel, which will allow you to manage and cleanse data more effectively. Snowflake is also close to enabling you to host Streamlit in Snowflake, under the Data Native App Framework, furthering their one data cloud goals.
Streaming + Dynamic Tables
Snowflake announced the debut of Dynamic Tables, now available in public preview. Dynamic Tables allow users to perform transformations on real-time streaming data, for example via the Snowflake Kafka Connector, which is near general availability. Dynamic Table transformations are defined with a SELECT statement, allowing for flexible transformation logic that is applied directly after the streaming data lands in Snowflake. It’s as simple as defining a view definition, but with the cost efficiency of a table, all with real-time streaming data.
As a Snowflake Premier Partner, OneSix helps companies build the strategy, technology, and teams they need to unlock the power of their data. Reach out to learn more about Snowflake’s latest innovations and how we can help you get the most out of your investment.