Intelligence Day held on the 23rd...LG Electronics and Hanwha share cases
Data Intelligence Platform, Data Lakehouse AI Capabilities Integration
Mosaic AI Acquisition Development...Open Source LLM 'DBRX' Released

Databricks is changing the paradigm of its products thanks to the generative AI trend. Databricks presented its 'Data Intelligence Platform' as an environment that enables customers to introduce an intelligence engine to their data lakehouse and optimize AI-based business operations.

On the 23rd, Databricks held a 'Data Intelligence Day' and a press conference at the COEX Intercontinental Hotel to introduce Databricks' data intelligence platform and 'DBRX', an LLM (Large Language Model) launched in March.

The core of the data intelligence platform is that it provides a new platform that allows for easy expansion and use of data and AI by adding a data intelligence engine to the data lakehouse architecture previously developed by Databricks.

If the data lakehouse provided the benefit of accessing all types of structured and unstructured data in one place, the explanation is that adding generative AI to it popularizes data and AI across the organization, ensuring cost efficiency and high performance in various workloads.

“Enterprises that struggle to connect and integrate data scattered across disparate devices can now do so 10x easier with the help of generative AI on Databricks’ data intelligence platform,” said David Meyer, SVP of Product at Databricks.

For example, a data intelligence platform can select the right machine based on past execution when leveraging massive Delta Lake data. It also helps companies develop AI applications by saving and modifying code within the pipeline when tables change, ensuring that it is not interrupted.

“It’s critical to enable every user in an organization to access data,” said Ad Renta, SVP and GM of Databricks APAC. “Databricks is committed to helping customers deliver AI-powered experiences.”

Databricks acquired 'Mosaic AI' to build its own data intelligence engine and developed the open source LLM 'DBRX'. Databricks' DBRX provides high accuracy, is easy to control, and provides complete ownership of the model and data. In addition, according to Databricks' data, DBRX outperformed the language understanding, programming, and mathematics performance of five LLMs, including Lamar 2, Mixtral, Grok-1, and ChatGPT 3.5, at a lower cost than competitors.

But that doesn't necessarily mean you have to use DBRX, says David Meyer, senior vice president. Meta's Rama 3 has some advantages over DBRX," he said. "Databricks' ultimate goal is to provide a platform environment that allows customers to use the generative AI model that best suits them, providing a variety of choices."

Databricks Korea Solutions Architect Team Leader Kyung-Woon Jang also asserted, “Databricks’ role is to provide an environment where customers can easily change models to fit their company’s scale.”

Team Leader Jang demonstrated how data developers and non-experts alike can leverage improved capabilities by incorporating AI into the data intelligence platform, including user UI, data indexing methods, and query planning methods.

Meanwhile, the Databricks event that day featured case presentations by domestic companies that have adopted the data intelligence platform. Databricks, which has 12,000 customers worldwide, has domestic customers such as LG Electronics, Hanwha, Shinsegae, and E-Mart 24.

“Databricks achieved 100% year-over-year growth in Korea last fiscal year, driven by the surge in demand for data and AI,” said Ad Renta, Senior Vice President. As of January 31st of this year, this amounts to approximately 1.6 billion dollars (approximately 2.2 trillion won) worldwide.

“By leveraging Databricks’ data intelligence platform, we have connected and integrated a vast amount of customer data, enabling thousands of users to simultaneously analyze customer data to provide AI-based products and services,” said Kwak Yong-hoon, head of LG Electronics’ Data Platform Office. “Through our partnership with Databricks, we will continue to engage in DX activities that can increase the value of customers who use LG Electronics’ products and services.”

Hanwha introduced Databricks to improve company-wide manufacturing efficiency and reduce development time by up to 90%. E-Mart 24 said, "We have over 80,000 products and 6 billion transactions and data, and after introducing Databricks, the time to derive insight was reduced by 96% and costs were also reduced by 93%."