German artificial intelligence startup deepset GmbH stepped into the limelight today, closing on a $14 million early-stage round of funding led by GV and announcing the beta launch of deepset Cloud, an enterprise software-as-a-service platform for building “language aware” software.

Today’s Series A round saw the participation of investors including Harpoon Ventures, System.One, Lunar Ventures and Acequia Capital, as well as the founders of leading machine learning and open-source companies such as Snorkel AI, Cockroach Labs Inc., Cloudera Inc., DeepMind, Neo4J Inc. and NGINX.

Deepset is the company behind the popular open-source Haystack project, which is a natural language processing framework that creates a layer for language awareness within the enterprise technology stack. With Haystack, developers can develop applications that allow users to access and interact with data using spoken commands. Haystack leverages so-called Transformer models and allows developers to build application programming interfaces to integrate voice-activated capabilities such as question answering, semantic search and summarization into their applications.

Deepset co-founder and Chief Executive Milos Rusic (pictured right, alongside Chief Technology Officer Malte Pietsch and Head of Machine Learning Timo Möller) said Haystack can be thought of as the “missing piece” between data and software applications, giving users a way to access, write and even create data using spoken language. “Deepset is helping [users] make machines understand language better, and we’re doing this for each and every software tool.”

Haystack, in other words, is the engine that powers all manner of voice-activated applications. The framework is integrated with tools such as Elasticsearch, an open search and analytics engine for all types of data, including textual, numerical, geospatial, structured and unstructured data, as well as Pinecone, a startup that has built a vector database for indexing and searching unstructured data.

Deepset Cloud, now available in beta, is the company’s first commercial offering. It’s essentially a fully managed version of Haystack that supports developers throughout the entire product lifecycle. Developers can use deepset Cloud to experiment, label and monitor new NLP services, prototype their applications and finally move them into production. The software is said to support large-scale business applications with many end users, and comes with collaborative tools for developers.

Rusic said he’s expecting a lot of interest in deepset Cloud, given the popularity of the open-source version of Haystack. Since its first release less than two years ago, Haystack has hit more than 100 contributors and is used by multiple enterprises in the Global 500.

“There is a high demand in NLP and companies look for frameworks and platforms that support them in building full-stack NLP services fast and fitting to their business applications,” Rusic said.

RedMonk analyst James Governor said that although natural language processing is increasingly being used for a wide range of applications, especially around enterprise search, it remains an esoteric discipline. “The open-source Haystack project is designed with practical adoption in mind, aimed at enterprise developers, so they can build NLP apps quickly, and with quality,” he said.

Image: deepset