Artificial intelligence platform vendor Immuta Inc. said today it has raised $20 million to build and sell its data access platform optimized for artificial intelligence model development in regulated environments.

The Series B round brings the company’s total funding to $29.5 million. It raised an $8 million Series A round last year. The new funding round was led by DFJ Growth Management LLC, with participation from new investors Dell Technologies Capital and Citi Ventures Inc., along with existing investors Drive Capital LLC and Greycroft Partners LLC.

The College Park, Maryland-based firm is tackling the problem of inefficiencies in data science and analytics created by the need for organizations to harmonize data from multiple sources while remaining compliant with a growing raft of regulations. It provides a unified data platform that lets data scientists, data owners and governance teams share raw data without the need to make copies and subject each to compliance scrutiny.

“The process we see enterprises currently using is designed around application-centric methodologies,” said Andrew Gilman, Immuta’s chief customer officer. “Managing data for data science is different.”

Specifically, machine learning and deep learning programs require access to data for use in building training models. Unlike traditional applications, which subject all data to the same set of algorithms, artificial intelligence programs process data differently depending upon its characteristics. That means that small variances in training data can have significant impacts on results. For example, information used in an automated mortgage approval process must be masked to eliminate race as a factor in order to meet anti-redlining rules.

Examining and preparing data for use in these training models can be an arduous process. In a typical scenario, an analyst must submit a request for data to the information technology organization, which makes a copy and scrubs it for compliance with regulations and internal governance rules, a process that can consume months. Copies of production data must also be subjected to security reviews, and consistency with production sources is never guaranteed once a copy is made.

Immuta’s read-only platform enables developers and data scientists to tap into a large number of supported storage sources, whether on-premises or in cloud, to read from live data and to join data from multiple sources in a single model. The platform has built-in governance features that mask or anonymize data to meet compliance guidelines. “Our ability to put laws into algorithms is our core intellectual property,” Gilman said. “You’re able to restrict the training set to only what you want, so the model is more accurate.”

The company emerged from stealth four years ago after incubating its technology in government projects. It logged its first sales outside of the government sector. Immuta has commercial customers in the life sciences, insurance and manufacturing industries. Gilman said.

Image: Unsplash