In 2020, Sridhar Ramaswamy co-founded Neeva with another former Google executive, Vivek Raghunathan. Neeva is an ad-free and privacy-focused search engine. In the last round of funding in 2021, Neeva was valued at approximately $250 million. Fast forward to the present, Snowflake is a well-known enterprise name. The company recently launched a series of open-source models, including Arctic LLM, designed for enterprises looking to create session SQL data copies, code copies, and RAG chatbots using large language models (LLM). This is thanks to Sridhar Ramaswamy, who became the new CEO of Snowflake earlier this year. Since taking office, the company has transformed from a pure data management service provider to a data and AI-driven entity, with a strong focus on generative AI. In a recent interview after taking the helm at Snowflake, Ramaswamy said, "I see this as a huge opportunity in the field of data applications and AI. It will keep me busy for many years to come." Baris Gultekin, Head of Snowflake AI, worked with Ramaswamy at Google for over 20 years and described him as an incredible leader. Gultekin said, "Sridhar brings incredible depth in AI and data systems. He has managed large-scale data systems and AI systems at Google." The expertise of Neeva in generative AI and LLM has now been integrated into Snowflake Data Cloud, enhancing Snowflake's AI capabilities, particularly in natural language processing and search functionality within its cloud data platform. Gultekin stated, "Neeva is an important acquisition for Snowflake. We are integrating many aspects of Neeva into Snowflake's products, with the most noticeable being Snowflake's universal search product." Universal Search helps customers quickly and easily find database objects, data products available in the Snowflake Marketplace, relevant Snowflake Documentation topics, and Snowflake-Community knowledge base articles within their accounts. Snowflake's generative AI project While there are several generative AI models in the market, Snowflake has chosen to target the niche market of enterprise customers. Recently, the company launched Snowflake Cortex. Cortex allows pre-trained LLMs from various vendors, including Snowflake's own Arctic LLM, to perform tasks such as text summarization, sentiment analysis, question answering, and code generation within the Snowflake environment. Furthermore, Cortex provides pre-built SQL functions that allow users to perform machine learning tasks on their data without extensive coding expertise. These functions handle tasks such as classification, regression, and anomaly detection. Snowflake also collaborates with Mistral, Meta, and Reka to host their LLMs on Cortex. Gultekin said, "We have partnered with Landing AI, AI21 Labs, and other capable partners to develop amazing products. They are important to us as they allow us to offer choices to our customers." Gultekin further stated that Snowflake is developing LLMs at a very affordable price while prioritizing the security of customer data. He said, "Despite a 17-fold reduction in the computational budget used, Arctic is on par with Llama 3 70B in terms of language understanding and reasoning." Additionally, he mentioned that they have 10,000 customers entrusting sensitive data to Snowflake. Considering this, he emphasized that all LLMs they operate are within strict security parameters, meaning no data is left behind and everything remains secure. Furthermore, he added that although Arctic LLM is several orders of magnitude smaller than OpenAI, the benchmark proves their excellent performance in document understanding and document data model question answering. Snowflake recently introduced Document AI, which extracts valuable content from unstructured data such as PDFs, images, and videos. It is supported by the multimodal large language model Arctic TILT, providing efficient content extraction for enterprises. Gultekin concluded, "We are just getting started. There is a lot to build. Our core use case is being able to have a conversation with data and how we make it better and easier." He mentioned that they recently released a bunch of products for public preview, including a series of chat products that can chat using structured data.