Introduction:
Deepset, a platform for creating business apps powered by extensive language models akin to ChatGPT, revealed today that it LLM-focused MLOps Offerings had raised $30 million in a funding round led by Balderton Capital, including GV and Harpoon Ventures.
According to co-founder and CEO Milos Rusic, the money will go toward enhancing Deepset’s offerings and expanding its staff, which currently numbers about 50 employees, to between 70 and 75 by the end of the year.
“Data science teams are still frequently the go-to resource for ‘all things AI,’ according to many corporations. In truth, many data science teams are restructuring, relearning, and changing their practices to meet the expanding expectations of the product teams and the end users in the organization, Rusic told TechCrunch in an email interview. “The industry is moving from AI labs to AI factories — it’s not anymore about tinkering, it’s about shipping successful products and value,” says the author.
Data science teams need to be more energized and overwhelmed, as Rusic implies. One recent survey found that the vast majority of data engineers, or data scientists who prepare data for analytics tools, are suffering from burnout, likely to switch employers within the next 12 months, and thinking about leaving the field entirely.
The terrible situation is probably making it more difficult for businesses to create AI. Only around half of AI initiatives get from pilot to production, according to a 2022 Gartner survey, and 53% of machine learning models are never used.
In 2018, Rusic, Timo Möller, and Malte Pietsch jointly formed Deepset, bootstrapping the company by building unique natural language processing models for businesses. The Transformer AI model architecture, created by Google in 2017, was closely followed by the three co-founders and served as the foundation for complex LLMs like ChatGPT and GPT-4.
The business introduced Deepset Cloud the previous year, which Rusic calls an “enterprise LLM platform for AI teams.” By offering a platform where users can test out various LLMs, integrate them into apps, distribute the applications and LLMs to end users, and conduct evaluations of the LLMs’ correctness while continuously tracking their performance, Deepset Cloud expands Haystack.
LLM-focused MLOps Offerings:
LLM-focused MLOps Offerings (Image Source: techcrunch.com)
Additionally, Deepset Cloud has elements for measuring and reducing LLM problems like hallucinations. Even the greatest LLMs today suffer from hallucinations, which lead to models fabricating erroneous information or truths that aren’t supported by actual events or data.
The primary opponents of Deepest, which has raised $46 million in funding so far, are companies operating in the MLOps market. By providing tools for each model’s life cycle stage, MLOps aims to make creating and managing machine learning models more efficient.
MLOps solutions, platforms, and services are offered to enterprise clients by an increasing number of startups in addition to market leaders like AWS, Azure, and Google Cloud. Seldon, which just received $20 million, and Galileo, Iguazio, which is owned by McKinsey, Diveplane, Arize, and Tecton are a few examples.
According to Allied Market Research, the MLOps market will grow from about $1 billion in 2021 to $23.1 billion by 2031. The magnitude of the addressable market will undoubtedly continue to draw new competitors.
Rustic, though, cites Deepset’s growth as proof that it is differentiating itself from the competition. The startup’s technology supports “hundreds” of customer pipelines, including workloads for Siemens and Airbus. Manz, a publisher of legal materials, collaborated with Deepset to introduce an internal AI-powered application that helps uncover court records, relevant precedents, and more. In the meantime, Airbus is utilizing Haystack to create apps that provide pilots in the cockpit advice on how to operate planes.
When building strong back-end application development consistently, Rusic said, “It’s frequently 10x faster to do so with Deepset Cloud than it is to hire, train, and manage a dedicated team.” Customers can employ many LLMs simultaneously with Deepset Cloud by integrating them into the application architecture to avoid vendor lock-in and mitigate data privacy and model sovereignty difficulties.
