Machine learning startup Indico Data Solutions Inc. has just been handed $4 million for its effort to upend the nascent world of artificial intelligence.

The seed funding round was led by Osage Venture Partners, with participation from existing investors .406 Ventures, Boston Seed, and Hyperplane.

Boston-based Indico aims to make machine learning more accessible to the masses of smaller organizations that have yet to adopt the technology. The issue, it said, is that machine learning’s potential remains largely untapped because of the enormous amounts of data needed to make it work.

Indico said most smaller organizations do not have access to the massive data sets needed to train machine learning models to achieve a sufficient degree of accuracy and reliability. Meanwhile, others lack the infrastructure and skill sets required to get the most out of machine learning.

The startup is employing a new approach to machine learning in order to get around these problems, Chief Executive Tom Wilde told SiliconANGLE. Indico’s specific approach is something called “Transfer Learning,” which focuses on storing knowledge or data gained while solving one problem, labeling that data, and then applying it to a different but related problem. For example, knowledge gained while learning to recognize cars could later be applied to a model that’s designed to recognize trucks.

Through transfer learning, Indico enables enterprises to create customized machine learning models using much smaller data sets, Wilde explained. The company makes a big claim about this, saying on its website that it can reduce the amount of data required to train machine learning models from hundreds of thousands of examples to just a few dozen.

“Indico has built a massive ‘generalized’ model on hundreds of millions of labeled data points,” Wilde said. “This model allows our customers to build their own small, but highly accurate custom models to solve their unique use cases. Our approach delivers high-performance models with only a fraction of the training data required.”

To begin with, Indico is using its accessible machine learning tools to tackle one of the most common problems that enterprises face, namely tapping into unstructured data and content such as text, images and audio files. The idea is that customers can use Indico to create machine learning models that quickly can dig into their data, thereby providing insights that help them to make better business decisions. Indico said its machine learning models can also be used to improve business workflow processes.

“Indico is addressing a really important issue for many enterprises – how to apply the benefits of artificial intelligence and machine learning to all the valuable, but often messy, unstructured data found in documents, text, images and audio,” said Osage partner David Drahms. “This type of content makes up 70 to 80 percent of the data in most organizations. The ability to automate document-based processes and more easily extract new insights from this content can unlock tremendous business value.”

Image: geralt/pixabay