In anticipation of the AI & Big Data Expo Europe, AI News sat down with Ivo Everts, Senior Solutions Architect at Databricks, to discuss groundbreaking developments in open-source AI and data governance.
One of Databricks’ key innovations is the DBRX model, which has set a new standard for open large language models (LLMs). Everts explained that DBRX outperforms other leading models, offering up to twice the inference speed of competitors like Llama2-70B. This efficiency is attributed to several technological advances, making DBRX a highly cost-effective option for organisations looking to train world-class LLMs.
“The goal is to democratise LLM training beyond a few providers,” said Everts. Databricks aims to show organisations they can achieve superior results with their own data in a cost-efficient way.
Databricks has also made strides in data governance by open-sourcing its Unity Catalog, which enhances data management across multiple cloud platforms and on-premise infrastructures. Unity Catalog helps organisations streamline data access and governance, offering features such as role-based access control (RBAC), data lineage tracking, and cross-cloud support.
“With Unity Catalog, companies can uniformly apply governance policies across multi-cloud and hybrid environments, simplifying data management,” Everts explained.
Databricks continues to push innovation with its new AI/BI product, designed to make business intelligence more intuitive through AI-enhanced data exploration and visualisation. The system includes:
- AI-powered dashboards for creating interactive visualisations with minimal coding.
- Genie, a conversational interface for natural language queries, which adapts to user needs and improves over time.
“This system allows users to gain deeper insights from their data, empowering self-service analytics across the organisation,” Everts added.
Another significant development is Mosaic AI, a comprehensive platform for building and managing machine learning and generative AI applications. Mosaic AI integrates enterprise data for improved governance and offers advanced tools for model deployment, monitoring, and cost-efficient custom LLMs.
“Mosaic AI provides a unified toolkit for deploying AI solutions, with support for prompt engineering and fine-tuning,” Everts said. Key components include centralised model management, real-time governance, and tools for predictive and generative AI applications.
At the core of these innovations is the Data Intelligence Platform, which combines the best of data lakes and data warehouses for real-time processing and secure data sharing. Powered by Delta Lake technology, the platform supports machine learning libraries like PyTorch and TensorFlow, offering scalability and performance through its cloud-native architecture.
Databricks is set to showcase these innovations at the AI & Big Data Expo Europe, highlighting how organisations can use the Data Intelligence Platform to create and deploy custom GenAI applications.
“We’ll demonstrate how to build a custom GenAI app using open-source models and data from Unity Catalog,” Everts said. The showcase will include live examples, such as generating cartoon images using the platform’s capabilities.
