TL;DR:
- Deepset, a platform akin to ChatGPT, raised $30M led by Balderton Capital with GV and Harpoon Ventures.
- Funds to expand products, services, and team from 50 to 70-75 members by year-end.
- Data science teams adapting to the growing demands of product teams and enterprise end-users.
- Burnout among data engineers affecting AI development; 53% of machine learning models never deployed.
- Deepset originated with NLP model training, evolved to Haystack, and recently introduced Deepset Cloud.
- Deepset Cloud offers versatile LLM experimentation, integration, deployment, and performance analysis.
- Challenges like hallucinations are addressed in Deepset Cloud for enhanced LLM accuracy.
- MLOps sector is expanding rapidly, projected to reach $23.1B by 2031, attracting new entrants.
- Deepset stands out with customer pipelines, partnerships with Siemens and Airbus, and unique offerings.
- Deepset’s growth indicates its distinctive position, aiding enterprise AI endeavors.
Main AI News:
In a significant move, Deepset, a groundbreaking platform specializing in crafting enterprise applications empowered by expansive language models akin to ChatGPT, has recently announced the successful conclusion of a $30 million funding round. The financing was led by prominent investor Balderton Capital, with notable participation from GV and Harpoon Ventures.
This substantial influx of capital is strategically earmarked for the amplification of Deepset’s portfolio of products and services, with a concurrent augmentation of its workforce. Currently comprising around 50 professionals, Deepset’s team is slated to grow to a robust 70 to 75 individuals by year-end, as shared by co-founder and CEO Milos Rusic.
Rusic’s insightful commentary underscores the evolving landscape within organizations where data science teams have traditionally been the pivotal force for all things related to AI. He asserts that this paradigm is shifting remarkably, as these teams adapt to the escalating demands of product teams and end-users in the enterprise realm. No longer confined to experimental realms, the industry is witnessing a transition from AI labs to AI factories, marked by a resolute focus on delivering tangible and value-rich products.
Indeed, Rusic is justified in his observation that data science teams are grappling with escalating workloads and burnout. Pertinent research reveals that a considerable majority of data engineers, who are instrumental in preparing data for analytical purposes, are undergoing burnout. This alarming trend indicates a high likelihood of these professionals seeking alternative employment opportunities within a year and even pondering an exit from the field altogether.
Regrettably, this scenario has potential repercussions on AI development within the corporate milieu. A recent survey by Gartner in 2022 disclosed that only around half of AI projects successfully transition from pilot phases to full-fledged production. Moreover, a staggering 53% of machine learning models fail to reach deployment.
The inception of Deepset by Milos Rusic, Malte Pietsch, and Timo Möller took place in 2018. The enterprise embarked on its journey by training bespoke natural language processing models for corporate entities. Drawing inspiration from Google’s transformative Transformer AI model architecture introduced in 2017, Deepset’s founders laid the groundwork for sophisticated LLMs (large language models) akin to ChatGPT and GPT-4.
Subsequently, the trio introduced Haystack in 2019, an open-source framework geared towards developing NLP back-end services employing Transformers and analogous LLM architectures. This initiative aimed to provide software engineers with a comprehensive suite of tools to swiftly create applications driven by LLMs, particularly those addressing specific use cases, such as aiding legal teams in exhaustive case file searches.
The evolutionary trajectory of Deepset eventually surpassed the confines of Haystack. In the preceding year, the startup unveiled Deepset Cloud, aptly termed an “enterprise LLM platform for AI teams.” This cutting-edge platform expands upon the foundation laid by Haystack, offering customers a platform to experiment with diverse LLMs, integrate them into applications, deploy the amalgamated solutions to end-users, and meticulously scrutinize the accuracy of LLMs while ensuring sustained peak performance.
Notably, Deepset Cloud is also fortified with components to gauge and mitigate prevalent issues linked with LLMs, notably hallucination. Even the most advanced LLMs are susceptible to this phenomenon, wherein models generate inaccurate or fictitious information devoid of any real-world basis.
Rusic elucidates that Deepset Cloud harnesses the Haystack technology extensively, furnishing an array of building blocks that negate the need for arduous, repetitive tasks. This empowers developers to concentrate on dispensing NLP back-end services that are API-driven, effortlessly composable, seamlessly embeddable, and subject to continuous monitoring.
Accumulating a total of $46 million in funding, Deepset perceives its primary rivals to be vendors operating within the MLOps space. MLOps endeavors to streamline the multifaceted process of conceptualizing and managing machine learning models by offering a gamut of tools that cater to every stage of a model’s lifecycle.
Besides established giants like AWS, Azure, and Google Cloud, a burgeoning cohort of startups now offer MLOps solutions, platforms, and services targeted at enterprise clients. Noteworthy players include Seldon, which recently secured a $20 million investment, as well as Galileo, Iguazio (owned by McKinsey), Diveplane, Arize, and Tecton, among others.
According to Allied Market Research, the MLOps sector is poised to burgeon to a staggering $23.1 billion by 2031, up from approximately $1 billion in 2021. The sheer potential of this market has led to a continuous influx of fresh contenders aiming to carve out their niche.
Rusic confidently points to Deepset’s trajectory as a testament to its uniqueness in the field. The startup boasts a plethora of customer pipelines currently operational on its platform, including substantial projects with Siemens and Airbus. Manz, a prominent legal publishing house, partnered with Deepset to launch an internal AI-powered tool designed to unearth court documents and related precedents. Airbus, on the other hand, has harnessed Haystack to construct applications that proffer aircraft operations guidance to pilots within the cockpit.
Rusic aptly encapsulates the value proposition of Deepset Cloud, emphasizing that it expedites the development of production-ready NLP and LLM services by a factor of ten when juxtaposed with the process of assembling and managing an exclusive team for back-end application development. By enabling customers to concurrently leverage various LLMs, Deepset Cloud offers a strategic solution to bypass vendor lock-in while effectively addressing concerns linked to data privacy and model sovereignty.
Conclusion:
Deepset’s significant investment exemplifies its prominence in the LLM-focused MLOps landscape. With its strategic funding and innovative solutions like Deepset Cloud, the company is well-positioned to navigate the challenges of AI development and contribute to the expanding MLOps market, which is projected to witness substantial growth in the coming years.
