Snowflake describes itself as a data platform and an AI data cloud.
“Sure, so snowflake is a data platform. We call ourselves an AI data cloud.”
Person
Baris Gultekin
1 episodes · 18 quoted claims
Neva was a web-scale search engine whose technology Snowflake brought into its search and RAG products.
“We acquired, um, a company that that I came with, uh, called Neva, which was a, uh, a a web scale search engine”
Baris Gultekin says 80 to 90 percent of all data is unstructured.
“you know, eighty to 90% of all data is unstructured data.”
Gultekin says real-world enterprise databases can have thousands of tables with hundreds of thousands of columns, which makes text-to-SQL difficult.
“we're talking about thousands and thousands of tables that that have hundreds of thousands of columns in them to be able to go and reason about.”
Gultekin says Snowflake Intelligence, its agent product for business users, is the fastest-growing product on Snowflake.
“and it is the fastest growing product that we have on Snowflake.”
Snowflake is working with Tableau, Omni and other technology providers on the Open Semantic Interchange, an open format for sharing semantic models between platforms.
“we're working with companies like Tableau, um, Omni, uh, you know, BBI platforms, as well as, uh, as well as others technology providers”
Snowflake built a fine-tuned document extraction model that is multiple orders of magnitude smaller than large language models, making it cheaper and faster for bulk extraction.
“we have a document extraction model that that we've, uh, build and fine-tuned. It is multiple orders of magnitude smaller than these large language models.”
Gultekin cites Cursor as an example of a company for which a custom autocomplete model makes sense at its scale.
“At their scale, it does make a lot of sense for them to have a custom model that is doing their, you know, autocomplete, for instance.”
Snowflake offers DeepSeek among the models available to its customers.
“We also have DeepSeek as a model that we provide for customers.”
Gultekin says that about two and a half years before the interview, Snowflake customers made clear they did not want data leaving Snowflake's security boundary.
“when we started the journey two and a half years ago or so, we heard loud and clear that our customers do not want to move their data out of the Snowflake security boundary.”
Gultekin says the models Snowflake uses act as sub-processors and save no state.
“So essentially, these models become sub processors. There is no, there is no state that's saved in any of these models.”
Gultekin says the cloud providers run model inference in their own stacks, while model providers keep ownership of the weights.
“the inference is run, uh, by, you know, the cloud providers in in in their stack.”
Snowflake built an internal sales assistant on Snowflake Intelligence and deployed it to 5,000 sellers.
“For instance, for the whole company, we've built a sales assistant and we've deployed to 5,000 sellers.”
Snowflake offers guardrail products that detect content such as hate speech and violence so the model can be stopped from responding.
“At Snowflake, we offer products to check for guardrails, they're doing things like checking for hate speech, violence, and other violations.”
Gultekin says Snowflake did not yet support the Agent2Agent (A2A) protocol and that customers were asking for agent-to-agent communication in general.
“still early, we don't yet support it. Uh, we're starting to hear our our customers don't necessarily ask for A to A specifically”
A Snowflake customer, identified in the transcript as 'SMP', analyzed earnings calls to see whether CEOs answered analysts' questions directly or indirectly, as a signal for stock selection.
“One interesting study, one of our customers as SMP, they analyze the earnings calls to understand when CEOs are responding to analysts questions in, um, either directly or indirectly”
Snowflake supports Apache Iceberg, an open file format for storing data.
“Uh, we support, uh, iceberg, which is a, uh, you know, open file format.”
Gultekin says tabular foundation models are an emerging trend that makes it possible to quickly build forecasting models.
“These tabular foundation models are interesting, being able to quickly build forecasting models and so forth are now possible.”