Pulls in another $72.5 million in funding
H2O.ai officials on Tuesday announced enhancements to its Driverless AI offering, an open-source platform aimed at making it easier for enterprises to adopt artificial intelligence (AI) and machine learning within their operations. In addition, the company said it has raised $72.5 million in a Series D round.
The Lowdown: The company’s platform automates many of the steps in an enterprise’s use of AI and machine learning, giving mainstream businesses capabilities similar to larger corporations and hyperscalers. The enhanced solution enables customers to create recipes to better customize the platform, new management and deployment capabilities, and a way to check for fairness and bias in machine learning.
The Details: Over the past year, the company had rolled out time-series and natural-language processing recipes that enterprises could use. Now businesses can create their own recipes to allow data scientists to customize Driverless AI through models, transformers, and scorers and extend the platform to better meet their individual needs. The customized recipes are treated as first-class citizens by the automated platform.
Other new features include the first set of vertical-specific solutions for anti-money laundering, customer 360, and malicious domain detection; 100 open-source recipes that customers can use; Project Workspace to enable data scientists to collaborate on disparate projects and deploy models in different environments; and a new module that monitors deployed models or system health. H2O.ai also added the ability to test machine learning models for sociological biases.
The latest round of funding was led by Goldman Sachs and the Ping An Global Voyager Fund, with continued investments from Wells Fargo, Nvidia, and Nexus Venture Partners. H2O.ai has raised a total of $147 million.
Background: H2O.ai’s Driverless AI platform is used by more than 18,000 companies and hundreds of thousands of data scientists. It’s used in a wide array of verticals that need to collect, storage, process, and analyze large amounts of data, including financial services, insurance, health care, retail, and pharmaceuticals.
The Buzz: “Every company needs to be an AI company. Our mission to democratize AI and empower all of our customers to be AI superpowers is one step closer with Bring Your Own Recipes in Driverless AI,” said Sri Ambati, founder and CEO of H2O.ai. “Domain experts can participate in the AI revolution that’s transforming every vertical. We have created over 100 open-source recipes that are design patterns for AI and curated by our grandmasters, data scientists, and domain experts.”
“We’re expecting a rapid adoption of AI in capital markets, as AI models started demonstrating ROI,” said Ediz Ozkaya, head of machine learning stats at Goldman Sachs. “H2O.ai is at the forefront of the space with Driverless AI, which enables us to inject our domain-specific AI capability into the platform in a consistent manner while protecting in-house IP and staying compliant.”
“H2O Driverless AI speeds up machine learning by automating our data science workflow. With the new recipe capability, we can extend and customize the platform to meet our needs, such as estimating the prepayment risk of underlying loans in fixed-income assets like mortgage-backed securities,” said Chris Pham, senior vice president of data management and data science at Franklin Templeton.
“We are pleased with the recipe feature added to H2O Driverless AI,” said Yan Yang, vice president of data science at Deserve. “We can be more creative in how we evaluate and serve those new to credit with the ability to customize and extend the platform to meet our unique needs.”
“In a regulated industry such as banking, the ability to explain what any model does is an absolute requirement. Decisions made by models must be not only sound but also fair,” said Agus Sudjianto, executive vice president and head of corporate model risk at Wells Fargo. “The team at H2O.ai has tackled this with machine learning interpretability and disparate impact analysis to detect bias and fairness.”