Deepchecks raises $14M to continuously validate machine learning models

Continuous machine learning validation startup Deepchecks Ltd has closed a $14 million seed funding round and today launched its platform to the public.

The round was led by Alpha Wave, with participation from Hetz Ventures and Grove Ventures.

Deepchecks was co-founded by Chief Executive Officer Philip Tanner (pictured, right) and Chief Technology Officer Shir Choleb (left). It was originally intended as a traditional machine learning operational platform. But they soon realized that the bigger opportunity lay in validating rather than building machine learning models.

Machine learning is all the rage today thanks to the rapid emergence of ChatGPT and other generative artificial intelligence models that are taking the tech world by storm. Suddenly, the focus shifted from long-term research projects to building real products that are regularly deployed in production environments, similar to how software is deployed and continually updated.

As Deepchecks explains, it took years of software development to establish the processes and tools needed to deliver high-quality applications on time. Companies today expect the same high-quality machine learning models to be deployed and updated on a regular basis, even though the technology is relatively new. Worse, machine learning development has more moving parts and is more opaque, making the process much more difficult to predict.

According to Deepchecks, the only way to regularly update high-quality machine learning models is to leverage lessons learned from software development and apply regular testing and validation. That’s why we’ve built a platform that helps MLOps teams visualize machine learning models at every stage, from development to deployment to operations to production.

Tannor explained that DevOps teams never deploy software that has not been systematically tested first. “But this is what is happening with machine learning applications today,” he said. “Deepchecks brings a new approach to his MLOps, improving models by adding validation to every step of the machine learning lifecycle.”

Deepcheck’s open-source platform allows users to reuse and tune specific components to comprehensively test machine learning models and the datasets used to train them. Its platform includes monitoring tools as well as production-grade root cause analysis tools accessed from a single, comprehensive user interface.

“Deepchecks introduces a community-driven MLOps framework that gives data scientists and developers to executives a clear picture of how machine learning applications perform from research to production. We will be able to do that,” Chorev said.

Andy Thurai, Vice President and Principal Analyst at Constellation Research Inc., said Deepchecks may be more focused, but it operates in a very crowded MLOps market, and the company’s ML testing, validation , monitoring and root cause analysis tools are already widely available. “However, his automated CI/CD for ML models works by testing whether models meet operational KPIs regarding performance, training test drift, and data integrity, and some users It could be an interesting option for us,” he said.

This is so far true, Deepchecks’ open source platform has been downloaded over 500,000 times by companies large and small, including Amazon Web Services Inc., Booking.com Inc. and Wix.com Ltd. Enterprise version available now, the platform offers additional security and collaboration features.

Photo: Deep Check