Exasol gets jolt of AI with Espresso suite of capabilities

A jolt of AI

AI, and generative AI in particular, has been a primary focus for many data management and analytics vendors over the past year.

OpenAI, a generative AI developer and vendor, launched ChatGPT in November 2022, marking a significant improvement in the functionality of generative AI large language models (LLMs).

Since then, tech giants such as AWS, Google and Microsoft have unveiled generative AI-fueled tools and expressed plans to make other generative AI features part of their roadmap. In addition, more specialized vendors such as ThoughtSpot and Tableau, which focus on analytics, as well as Dremio and Informatica, which focus on data management, have similarly introduced generative AI capabilities.

At the core of their AI-focused product development plans are increased efficiency and more widespread use of data.

Data management and analytics platforms are generally complex, requiring users to know code to query and analyze data. In addition, it is time-consuming to write all the code required to integrate tools, build data pipelines that combine data coming in from different sources and develop data products that enable analysis.

Both traditional AI and generative AI can change that.

LLMs enable true natural language processing (NLP), which greatly reduces the need to write code to query data, as well as translate text to code so that engineers can save time. Meanwhile, traditional AI and machine learning can augment LLMs by making them smarter over time, improving the accuracy of generative AI tools that without extensive training on relevant data can often deliver incorrect and misleading responses.

Now Exasol is among the data management vendors adding new AI capabilities.

The integration with AI code optimization vendor TurinTech's evoML is designed to help customers be more efficient by deploying machine learning models, including LLMs, directly in their database. Once within their database, users can run their models on big data by taking advantage of distributed and parallel processing engines that won't cause overloads that slow workloads.

Espresso AI complements the core in-memory engine with AI capabilities designed to enhance support for AI workloads and lower barriers to adoption of AI.

Exasol AI Lab is a new container-based tool that includes integrations with data science platforms, including Amazon SageMaker, Azure ML, Hugging Face, Ibis, Jupyter, PyTorch, scikit-learn and TensorFlow. Data experts can use Exasol in concert with their preferred development ecosystems.

Finally, Exasol's integration with Veezoo, a natural language query specialist, enables customers to ask questions of billions of rows of data using natural language and quickly get answers. That includes suggestions for follow-up questions that might lead to deeper insights.

The NLP capabilities resulting from the integration with Veezoo, in particular, are an important addition, according to Aslett.

He noted that nearly half of respondents to USG's 2024 "Market Lens AI Study" cited lack of skills and expertise as the biggest challenge to AI development and deployment.

"In addition to making it easier for data scientists to incorporate Espresso with their current tools and platforms, Espresso AI also tackles one of the biggest challenges related to adoption of AI," Aslett said. "Natural language query … can lower barriers to adoption and increase efficiency by translating natural language questions into database queries"

In fact, more than a third of the respondents to ISG's survey said they are already starting to use generative AI, he added.

With more enterprises seeking ways to build traditional AI and generative AI models and applications, customer feedback was one of the motivating factors behind Espresso AI's development, according to Mathias Golombek, Exasol's CTO.

Organizations are striving to eliminate data isolation to bring AI and analytics together, and Espresso AI was designed to do just that, he noted.

"Espresso AI is a solution … to bring AI and business intelligence more closely together," Golombek said. "We are in close contact with our customers, and quite often data science teams have been working in silos for many years. But during the last two years, there has been a significant shift as AI technology has become more important, more powerful and more accessible."

Future plans

One of the key pieces of Espresso AI is the AI Lab that integrates with a variety of data science platforms, according to Golombek.

As a result, further integrations are a significant part of Exasol's roadmap, he noted. While the initial integrations are with platforms from established vendors, the next set of integrations will be with open source tools.

"Exasol will continue to enhance its AI Lab to support adjacent data science tools, especially open source technologies," Golombek said. "Users will be able to create an environment where data scientists feel comfortable leveraging the maximum amount of data they can by running ML models directly within the Exasol database."

Aslett, meanwhile, said that Exasol would be wise to continue adding features that help customers develop generative AI models and applications.

While some data management vendors have developed capabilities such as vector search and retrieval-augmented generation (RAG) that enable customers to build and maintain generative AI models and applications, Exasol is just getting started with integrations that allow users to work with LLMs.

"Exasol's support for generative AI is comparatively nascent," Aslett said. "We anticipate seeing further details about support for vector search and RAG, which can be used to improve trust in the output of LLMs by augmenting foundation models with proprietary data and content from enterprise databases."

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.