Snowflake Cortex allows you to start building production ready GenAI solutions like chat with your own data, Document understanding, or use the Cortex functions to enhance your data with AI/ML. Cortex toolset is brand new, but we at Kaito already actively crafts solutions with our customers. Here you can read a bit about what we have already been working with.
Snowflake Cortex allows us to implement advanced document search features and combine them with GenAI models. GenAI models alone seldom work well in enterprise use-cases because more specific background information is needed for useful answers.
With Cortex we can give the solution ability to search company specific information from different sources and serve this information to the GenAI model together with for example the user question. Common use cases are Chatbots and customer service systems.
This approach is called RAG or retrieval augmented generation. Before Cortex RAG architectures needed group of separate technologies to work, leading to sensitive data moving along between many different systems.
With Cortex Kaito has built end-to-end RAG solution on top of Snowflake. Data stays inside Snowflake and solution is much easier to govern and maintain.
With Cortex Search we can handle the end-to-end process of data ingestion, embedding, retrieval, reranking and generation. Snowflake hosts most of the leading open-source large language models (LLM) and soon we can also fine-tune the models for specific use-cases using Cortex Fine-Tuning.
Snowflake Cortex Search & Simplified RAG Architecture
We have been working with one of our large manufacturing customers in making sense of complex technical documents using Snowflake Cortex. Using the Cortex Search it is easy to build a search optimised database from complex and varied documents.
Using Cortex we can recognise similar existing cases and summarise and categorise information etc. This allows our customer to automate and streamline processes reducing the workload of highly skilled technical experts.
There is a suite of ready made functions in Cortex that can be easily utilised as part of data processing and modelling workflows. These simple SQL functions allow us to enhance data in varied ways. From language translation, data cleaning to sentiment analysis.
Cortex ML functions make it possible to do simple analysis on time-series data like anomaly detection and forecasting. When the simple ready made functions are not enough it is possible to move towards more serious ML with Snowpark.
"We give meaning to information."
