UPDATED 09:00 EDT / JUNE 14 2021

AI

AI unicorn Dataiku takes its machine learning platform to the cloud

Dataiku Inc. today introduced a managed cloud edition of its artificial intelligence development platform that promises to reduce the amount of work involved in building custom machine learning models. 

New York-based Dataiku is backed by nearly $250 million in funding from investors including Alphabet Inc.’s CapitalG venture capital arm. It achieved unicorn status in 2019 after a secondary financing round.

Dataiku’s AI platform provides a set of tools that companies can use to create machine learning models and perform related tasks such as preparing training datasets. Until now, running the software required customers to set up a customized infrastructure environment on-premises or in the public cloud. Dataiku Online, the new managed edition that the startup introduced today, removes the need to maintain infrastructure by providing its platform as a cloud service. 

The process of building an AI model consists of two main steps: finding a machine learning algorithm suitable for the target use case and refining the algorithm by having it process large amounts of training data. Dataiku’s platform offers users a choice in how they carry out the task: They can create their own custom algorithm or have the platform choose one automatically using a feature dubbed AutoML.

The AutoML feature works by examining what kind of data a company wishes to process with its AI software and finding an algorithm that can carry out the processing efficiency. The algorithm is selected from a built-in library of AI models. From there, Dataiku’s platform refines the AI model using training data provided by the user.

Savvy enterprises sometimes opt to build custom AI algorithms for their applications. To support such projects, Dataiku offers tools that can be used by data scientists to create a neural network with custom configuration settings, or hyperparameters as they’re known in AI parlance. The process of finding the combination of settings that maximizes a neural network’s accuracy and performance for a given use case is known as hyperparameter optimization.

Dataiku enables data scientists to perform optimization using one of several methods. They can use random search, an approach that employs randomized trial and error to guess the optimal configuration of an AI. Or they can use grid search, a technique that is based on the same basic concept but involves testing a more limited number of potential AI configurations. Narrowing down the number of possible setting combinations reduces the amount of time it takes to go through the entire list, which in turn speeds up AI development.

Dataiku also supports a third, more complicated tactic for optimizing AI models’ configuration, known as Bayesian search. The technique involves bringing a second AI model into the loop that tries out different configurations and, using mathematical techniques, finds the areas of the so-called parameter space where it has the highest chance of stumbling upon the optimal mix of hyperparameters. It then proceeds along this path until finding the best match.

Deploying an AI in production involves numerous other steps as well, such as removing errors from the training set being used to fine-tune the model and monitoring the model’s accuracy to ensure it doesn’t become less efficient over time. Dataiku has built features for those tasks as well, which it’s bringing over to its newly announced cloud-based managed offering.

Besides reducing the amount of work involved in maintaining deployments of the platform, Dataiku hopes the managed offering will expand its addressable market. Currently, the startup’s platform is mainly used by large enterprises such as General Electric Co. and Unilever PLC. Now that the software is available in a form that doesn’t require setting up or maintaining infrastructure, it should become a more viable option for small and midsized firms.

“We started developing Dataiku Online to address the needs of small and midsized businesses, in addition to startups, who don’t rely on on-premise technologies or custom clouds the way our enterprise customers do,” explained Dataiku Chief Executive Officer Florian Douetteau.

For small startups that were founded less than five years ago or have raised under $10 million in funding, Dataiku is offering a separate edition of Dataiku Online with discounted pricing. Small software deals with early-stage startups can grow into much larger contracts over time. In the fast-paced technology market, it’s common for a well-funded new company to grow from a handful of staffers to dozens or hundreds of employees within a few years. 

Image: Dataiku

A message from John Furrier, co-founder of SiliconANGLE:

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU