Qwak, an all-in-one MLOps platform to build and deploy fashions, raises $12M

Israeli startup Qwak, which supplies enterprises with an end-to-end MLOps platform to build and deploy fashions at scale, at present introduced $12 million in a recent spherical of funding. The firm plans to use the capital to additional develop its product and ultimately arrange a “machine-learning cloud” for enterprises.

While machine studying (ML) has been a speaking level for a very long time, the yr 2022 noticed it go mainstream with the launch of generative AI purposes like Dall-E, MidJourney and ChatGPT. Enterprises at present are aggressively racing to build ML fashions to unlock worth throughout capabilities, be it real-time buyer help, fraud detection or defining a pricing technique.

>>Follow VentureBeat’s ongoing generative AI protection<<

However, when it comes to really constructing high-performing fashions and integrating them into merchandise, issues get difficult. Data science groups have to cope with a extremely fragmented surroundings the place they’ve to combine with totally different stakeholders like DevOps and information engineers and make the most of specialised instruments to build a easy mannequin pipeline. This takes up important time and sources – to the purpose the place many tasks don’t even make it to manufacturing. And, for the few fashions that make it to the manufacturing stage, deployment can take a very long time, adopted by the speedy want to consistently monitor them for high quality and effectivity.

Alon Lev, who beforehand led because the VP of information at Payoneer, noticed comparable challenges and discovered that solely the most important and most superior firms had the sources to build their very own inside ML platforms. The remainder of the business struggled to effectively flip concepts into ML fashions. This led him and fellow cofounders from AWS, ironSource and Wix to launch Qwak as a unified MLOps platform.

How does it work?

As Lev defined, Qwak integrates all elements of the MLOps life cycle in a single place, permitting the info science crew to function independently, proper from the stage of constructing the fashions, evaluating efficiency and analyzing modifications to transferring them to the manufacturing surroundings and driving monitoring efforts.

The platform is totally managed (hosted both on Qwak’s or the client’s cloud), which signifies that information science groups don’t want to set up packages or keep infrastructure, and the product takes care of all of the operational infrastructure. 

“At the end of the day, Qwak allows data science teams to be more effective, and to significantly shorten the model development time. Instead of many months, the whole process can be cut down to a few hours, allowing teams to iterate faster and improving quality testing of the ML models and their behavior,” Lev famous.

Since its launch in December 2020, Qwak claims to have witnessed 10-fold year-on-year progress with dozens of enterprises signing up for its platform, together with NetApp, Lightricks, Yotpo, JLL, Guesty and OpenWeb.

Competition in MLOps

The MLOps area has grown considerably with a number of open-source instruments and distributors wanting to assist enterprises build and deploy production-grade fashions, together with Deci, Domino Data Labs and H2O AI.

Qwak, for its half, claims to differentiate from these gamers by providing all of the elements and integrating them collectively.

“While there are many [vendors] that cover various components of Qwak — including feature store, model registry, serving, monitoring and ML pipeline orchestrators — the real power lies in creating a unified platform where all these parts are seamlessly integrated. By doing this, we provide a streamlined experience for data scientists, eliminating the friction of connecting multiple tools every time a model needs to be built or upgraded,” Lev famous.

This additionally improves visibility and facilitates the sharing of ML elements between crew members, enhancing collaboration and boosting productiveness, he added.

With this spherical, which was led by Bessemer Venture Partners, the corporate will proceed to build out this all-in-one providing and transfer towards its long-term imaginative and prescient of constructing a complete machine-learning cloud. It additionally plans to increase its crew within the U.S. and European markets. 

VentureBeat’s mission is to be a digital city sq. for technical decision-makers to acquire data about transformative enterprise know-how and transact. Discover our Briefings.