Deepset Secures $30 Million Funding to Revolutionize Enterprise AI Applications
Deepset, a platform designed to create enterprise applications powered by large language models similar to ChatGPT, has announced a successful funding round, raising $30 million. This round was led by Balderton Capital, with contributions from GV and Harpoon Ventures. The funds will be utilized to enhance Deepset's suite of products and services, as well as to expand its team from approximately 50 to between 70 and 75 employees by the end of the year, according to co-founder and CEO Milos Rusic.
“In many organizations, data science teams are still seen as the primary resource for all AI-related tasks. However, data science teams are undergoing restructuring and adapting their methodologies to meet the rising demands from product teams and end-users within enterprises,” Rusic explained in an email interview. “The landscape is shifting from AI labs to AI production lines — it’s no longer about experimentation but about delivering successful products and tangible value.”
Rusic's observation about the heavy workload facing data science teams is supported by recent research, which indicates that a significant percentage of data engineers—the professionals who prepare data for analytical tools—are experiencing burnout. Many are likely to seek employment elsewhere within a year or even consider leaving the industry altogether.
This challenging situation is likely exacerbating hurdles in AI development across enterprises. A 2022 survey from Gartner revealed that only about half of AI projects transition from pilot to production, with 53% of machine learning models never being deployed.
Founded in 2018 by Rusic, Malte Pietsch, and Timo Möller, Deepset initially bootstrapped by crafting custom natural language processing models for various enterprises. The co-founders closely monitored the development of the Transformer AI model architecture by Google in 2017, which laid the groundwork for advanced large language models like ChatGPT and GPT-4.
In 2019, Rusic and his co-founders launched Haystack, an open-source framework aimed at building NLP back-end services utilizing Transformers and other LLM architectures. The intention was to provide software engineers with a toolkit to swiftly create LLM-driven applications, particularly for specialized tasks like aiding legal teams in searching through case files.
However, Deepset’s ambitions quickly surpassed what Haystack could offer.
Last year, the company introduced Deepset Cloud—an “enterprise LLM platform for AI teams.” Rusic describes Deepset Cloud as an evolution of Haystack, enabling customers to experiment with various LLMs, integrate them into applications, deploy them to end users, and analyze their performance and accuracy consistently.
Deepset Cloud also incorporates features for assessing and addressing common challenges associated with large language models, including the issue of hallucination—when LLMs generate fabricated information or data not grounded in reality.
A visual representation of Deepset Cloud, its innovative MLOps platform.
“Deepset Cloud utilizes the Haystack technology extensively—its pipeline architecture, core components, datastores, and integrations,” Rusic detailed. “Our platform provides all the necessary building blocks, allowing developers to focus solely on deploying NLP back-end services that are API-driven, easily composable, embeddable, and monitored efficiently.”
Having raised a total of $46 million in funding to date, Deepset considers vendors competing in the MLOps (Machine Learning Operations) space as its primary competitors. MLOps strives to simplify the process of developing and managing machine learning models by equipping users with tools for each phase of a model's lifecycle.
In addition to established players like AWS, Azure, and Google Cloud, a growing number of startups in MLOps are catering to enterprise clients. Notable names include Seldon, which recently secured $20 million; Galileo; McKinsey-owned Iguazio; Diveplane; Arize; and Tecton.
According to Allied Market Research, the MLOps sector is forecasted to grow from approximately $1 billion in 2021 to $23.1 billion by 2031. The vast market potential will undoubtedly continue to draw new players into the field.
Rusic highlights Deepset’s expansion as evidence of its differentiation in a crowded market. The startup currently manages “hundreds” of customer pipelines, including projects for high-profile clients like Siemens and Airbus. Moreover, legal publishing company Manz has partnered with Deepset to create an AI-powered internal tool that efficiently identifies court documents, related precedents, and more. Airbus is utilizing Haystack to develop applications that provide operational guidance to pilots in-flight.
“It’s often ten times more efficient to repeatedly build production-ready NLP and LLM services using Deepset Cloud than to hire and maintain a dedicated team for robust back-end application development,” Rusic stated. “Deepset Cloud enables clients to leverage multiple LLMs concurrently, integrating them seamlessly into their application architecture to avoid vendor lock-in while addressing data privacy and model sovereignty concerns.”
Ranking
