Until now, enterprises and their employees have had to rely on basic keyword-based searches when trying to find information from their unstructured data — document collections, reports, insights and other textual information. In 2015, enterprises created 30% of the total data worldwide and by 2025, this share is expected to double. For companies sitting on a mountain of data, searching, retrieving, summarising, discovering and analysing textual data quickly and accurately is one of the biggest challenges they face and only increasing. LLMs, which are machine learning models capable of applying their learnings in seconds to solve a variety of business problems, could solve this.
In this context, deepset, an enterprise platform for building applications with LLMs, has raised €27 million in funding led by Balderton Capital. The round also included participation from existing investors GV (Google Ventures), and Harpoon Ventures.
deepset solves one of the biggest challenges facing enterprises today – how to exploit their own data more quickly, more easily and at scale. It does this by helping organisations tap into the potential of today’s most advanced and promising technology, large language models (LLMs). Enterprises who don’t have full expertise or time to build their own LLM platform can use deepset Cloud to harness the power of natural language processing (NLP) to build their own LLM-enabled applications.
Milos Rusic, co-founder of deepset, said: “Enterprises can see huge benefits from leveraging LLM technology. At deepset, we’re providing a platform that helps to bridge the decades of research in machine learning and computer science into production-ready applications. In the same way you don’t need to know much about microchip architecture to write software, you don’t need to be an NLP or LLM scientific researcher to use our Haystack framework and deepset Cloud.”
deepset Cloud can solve the organisational problems inherent in building bespoke LLM-powered applications by assisting AI engineers, software engineers and product managers to adapt rapid iteration methodology to NLP and consequently develop applications in a standardised, predictable manner. Since deepset Cloud makes it easy for the product team to establish a proper software development life cycle it can reduce time-to-market dramatically and help organisations to stay model-agnostic.
deepset Cloud has helped European legal publishing house Manz, for example, to access a new and fast-growing category of LLM-enabled products for legal professionals, without having ML specialists. The legal team can now instantly find precedents, relevant regulation, templates, cross-reference data, summarise findings, and more from millions of documents. This involves a significant time saving for anyone in legal, while allowing Manz to keep its data up-to-date and searchable, which ultimately helps the company to grow its customer base and stay relevant. Meanwhile, aircraft maker Airbus’s R&D team is using deepset’s open source technology, Haystack, to build an application that helps pilots discover and use most relevant aircraft operation guidelines from the cockpit, where time-to-information is of the absolute essence.
James Wise, partner at Balderton, commented: “The pace of change in the power of LLMs is remarkable, but to date there has been a lot of testing without the tools needed to get applications built of these new models into production. deepset Cloud and Haystack provide the tools and platform for enterprises to develop powerful new AI applications transparently and securely.”
deepset’s founders Milos Rusic, Malte Pietsch and Timo Möller first spotted the potential of natural language processing to transform enterprise back in 2018. Having seen barriers to building LLM apps first-hand, the team wanted to bridge the gap between the potential LLMs can bring to enterprises, with the fast-paced and complex nature of the technology and research.
This latest funding round takes the total raised by the Berlin-based firm to €42 million. deepset has also secured backing from experienced founders in open source, including Alex Ratner (Snorkel AI), Mustafa Suleyman (Deepmind), Spencer Kimball (Cockroach Labs), Jeff Hammerbacher (Cloudera) and Emil Eifrem (Neo4j).