Machine learning observability startup Aporia Technologies Ltd. said today it has raised $25 million in an early-stage round of funding to help enterprises trust the accuracy of their artificial intelligence models.

Today’s Series A round was led by the venture capital firm Tiger Global, with participation from Samsung Next and existing investors TLV Partners and Vertex Ventures. The raise comes less than a year after Aporia closed on a $5 million seed funding round, bringing its total funding to date to $30 million.

Aporia has created what looks to be an extremely useful observability platform for machine learning models. The idea with it is to empower businesses to trust their AI by providing full visibility into how their models are performing in the real world. Aporia’s platform helps teams to create customizable monitors that detect for real problems with AI models, such as biased predictions, unexpected changes in the format of its input data or degradation of performance over time.

Aporia’s platform goes further too, providing insights that can help data scientists investigate and understand what’s causing any issues that show up.

The most recent new capability of Aporia’s platform, unveiled today, helps to explain how machine learning models are actually making their predictions. That’s useful, Aporia said, because it can be difficult for business leaders to trust AI models with critical decisions.

Moreover, in some industries there may be regulatory requirements that demand companies explain how certain decisions are reached, for example with loan applications. So a financial services firm using AI needs to interpret how its models come to decisions.

Aporia says a deep understanding of AI is necessary, because if models go awry they can have a devastating impact on a business. Perhaps the most obvious illustration of this is Zillow Inc.’s failed house-flipping business model, which resulted from the failure of its AI to forecast house prices accurately, leading to millions of dollars in losses and thousands of workers being laid off.

Aporia co-founder and Chief Executive Liran Hason said visibility into the accuracy of machine learning models is a must-have for every organization that uses them. Building a tailor-made system is tricky though, hence the need for a platform like Aporia, he said.

“We see incredible demand for our platform as business leaders understand that AI isn’t a novelty or even a competitive advantage, it’s a now a necessity,” Hason continued. “AI has gone mainstream and just like software, AI has issues and bugs. These issues are now a daily reality that companies are grappling with and that are affecting all of us.”

The need for a platform such as Aporia’s was highlighted by Andy Thurai, an analyst at Constellation Research Inc., who told SiliconANGLE that the fast-track of AI adoption in enterprises has created lots of issues with models in production.

“Data engineering teams struggle with model drift, model bias, and data integrity issues,” Thurai said. “A properly built AI/ML infrastructure and model delivery pipeline should include tools to help with AI model data versioning pipeline, training orchestration, feature store, model serving, model monitoring, and model auditability among other things. If any one of them is not properly set up, the entire AI pipeline from getting the AI model from lab to production can and will fail.”

Thurai added that there’s big demand for solutions such as Aporia’s, because existing observability tools that monitor enterprise systems, storage, networks and hardware can’t really look into AI models.

“Model skew can happen as models drift and become less potent over time,” he explained. “Depending on the applicable area, it can happen either in a matter of days or months. Thus, it is important not to let a model go stale, so the decisions can be accurate.”

It’s no surprise then, that Aporia said it has seen a 600% increase in customer adoption over the last six months, with tech brands such as Armis Inc. and New Relic Inc. and the insurance firm Lemonade Inc. all using its platform.

With its new funding, Aporia intends to triple its team’s headcount over the next year and expand into the U.S. market, while increasing the number of use cases its platform can address.

“Aporia has demonstrated unbelievable growth since its launch and has amazing momentum, quickly becoming a leader in the space of ML observability,” said Tiger Global Partner John Curtius.

Photo: Aporia