Predibase debuts low-code machine learning platform for building generative AI models

Machine learning startup Predibase Inc. today announced the commercial availability of its low-code declarative machine learning platform for artificial intelligence developers, with new capabilities for large language models.

The launch of the company’s platform comes after the company announced a $12.2 million expansion to its Series A funding round led by Felicis Ventures Management Co., LLC.

Predibase aims to help developers and data scientists build, iterate, and deploy advanced AI models and applications. The company says it aims to help small businesses compete with big players like Apple, Meta Platforms and Uber Technologies. It does so by removing the need to use complex machine learning tools or construct low-level machine learning, they argue. Framework.

With Predibase’s machine learning platform, teams simply define what they want to predict using a large selection of pre-built AI models, and the platform does the rest. Novice users can leverage the recommended model architecture to get started, and experienced practitioners can use the platform to fine-tune their model parameters.

In this way, Predibase claims it can reduce the time it takes to deploy machine learning models from months to just a few days. More than 250 models have been trained on the platform since returning from stealth mode last year.

With the rise of generative AI models like OpenAI LP’s ChatGPT captivating the imagination of nearly every company today, the company couldn’t have picked a better time to launch its platform. In recent months, companies have rushed to implement generative AI capabilities to gain a competitive advantage over their competitors.

Predibase co-founder and CEO Piero Molino told SiliconANGLE that companies are desperate to build machine learning capabilities into their internal and customer-facing applications. The problem, he said, is that most machine learning development tools are too complex for engineering teams to use, while data science resources are too scarce. Therefore, Predibase uses what we call a “declarative” approach to machine learning development.

“Declarative means that you can specify what you want the ML model to predict and from what data you want it to predict without specifying how,” explained Molino. “In practice, this is a configuration YAML that matches the schema of your data and declares what you want the model to predict, rather than writing thousands of lines of low-level machine learning code to accomplish the same thing. It means writing a few lines of a file… Think about what Terraform does for your infrastructure.It’s the same approach applied to machine learning.”

According to Molino, Predibase’s mission is to make it easier for beginners and experts alike to build machine learning applications, including powerful, large-scale language models for generative AI applications, and deploy them in production. said it is. To do this, users can build on top of open-source LLMs such as Ludwig and Horovod, which are continuously developed and improved by the community, and fine-tune those models to suit their needs.

“The problem with the open source model is that companies need to understand how to deliver the open source model, how to adapt it to their tasks, and how to implement it in a cost-effective manner,” Molino said. I’m here. “Predibase addresses all three of these needs by fine-tuning and deploying by simply writing a simple declarative YAML configuration that any developer can write.”

The latest version of Predibase also features a new AI-powered data science copilot tool that provides developers with recommendations on how to improve the performance of their models under development, as well as explanations and examples. provided in real time.

unique approach

Andy Thurai, Vice President and Principal Analyst at Constellation Research Inc., told SiliconANGLE that Predibase’s declarative approach to machine learning is unprecedented and unique.

“Rather than building a model from scratch, users can quickly build the underlying workflow, architecture and tools to set up an experimental environment,” he said. “Predibase combines this with low-code options to enable rapid deployment of models to production using a variety of templates. Bringing machine learning to the masses rather than relying on costly data scientists For companies that lack good data science resources or need to build and deploy models faster than existing data science teams, this could be a good option. there is.”

To prove its worth, Predibase announced a two-week free trial of its platform, giving all companies the opportunity to see how its declarative approach can accelerate model development. Free trials are available as a fully hosted Software-as-a-Service via Predibase Cloud or a virtual private cloud in your own environment. As part of the trial, customers will have access to LudwigGPT, a custom LLM that powers Predibase’s own data science copilot.

The platform has already been extensively tested by a number of companies during its beta stage. Dr. Volkmar Schaaf Katz, head of data science at Wells Fargo, is one of its biggest proponents. He said the platform combines the simplicity of his AutoML platform with the robust flexibility and advanced features his data scientists demand. “It’s amazing how quickly we can deliver accurate results and cut time to value from months to days,” he said. “Furthermore, Predibase will allow different personnel to operate this platform for many use case scenarios in regulated areas such as finance and healthcare.”

The new investment from Felicis brings Predibase’s Series A funding round up to $25.2 million, bringing its total funding to date to $28.5 million, the company said. Predibase said it will use the additional funds to expand its go-to-market business and develop new platform features.

Image: Predibase