Galileo Technologies Inc., an enterprise artificial intelligence observability and evaluation platform provider, today announced it has raised $45 million in new funding.

The Series B round was led by Scale Venture Partners to help companies build trustworthy AI platforms. Databricks Ventures, Premji Invest, Amex Ventures, Citi Ventures, ServiceNow and SentinelOne participated alongside existing investors Battery Ventures, Walden Capital and The Factory. In addition, Andy Vitus, a partner at Scale Venture, will join Galileo’s board.

Large language models and the applications built on them depend on accurate responses. Today’s LLMs are becoming larger, more complex and more powerful than ever and that means going from a demo to production-ready is not easy while maintaining that accuracy. Galileo has created an all-in-one evaluation and observability stack that provides metrics and protection against common issues.

“We started Galileo three years ago to solve AI’s measurement problem, specifically with a focus on language models,” said co-founder and Chief Executive Vikram Chatterji. “Using humans or LLMs to judge model responses is expensive, slow and does not scale. Yet today these are the de-facto techniques adopted across AI teams.”

Chatterji argues that Galileo differentiates itself with a research-backed approach and a specially designed user experience that assists enterprise users manage LLM and LLM app evaluation and trust at the scale they need. According to Gartner Inc., the advancing rate of adoption means that more than 80% of enterprise businesses will include generative AI by 2026.

The company says AI has a “measurement problem,” where getting proper metrics on AI performance is difficult and requires specialized tooling. To handle this the company developed what it calls the Evaluation Intelligence Platform that embeds evaluations into the AI development pipeline so that teams can rigorously test LLMs, train and fine-tune, debug during experimentation, and secure after deployment.

This became the foundation of the company’s flagship product the Luna Evaluation Suite, a collection of high-performance evaluation foundation models designed to be accurate, fast and cost-effective. Out-of-the-box they can evaluate the performance of LLMs such as OpenAI’s GPT-4o and Google LLC’s Gemini Pro, requiring no ground truth. They can perform a large number of tests and gather a variety of insights without needing to prepare cumbersome test sets.

“Traditional evaluation methods, such as human evaluations or using LLMs as judges, fall short for enterprise use cases,” said Jim Nottingham, senior vice president and division president of advanced compute solutions at HP Inc. “Galileo’s rapid innovation and focus on overcoming the biggest evaluation hurdles — from accuracy to bias — provides a complete view of how gen AI apps are performing, which is why they’ve become a critical part of our Z by HP AI Studio.”

Given the increasing adoption of AI language models, Galileo said that since the beginning of 2024, the company has brought on six Fortune 50 companies as customers, including Comcast Corp. and Twilio Inc.

Chatterji said that the new funding will go toward deepening Galileo’s development and engineering efforts to meet market demands for more accurate and trustworthy AI systems and expand its product line.

Image: Galileo