Fiddler AI Gets $32M in Series B Funding to Grow Its Modeling Performance Management Platform
Since 2018, Fiddler AI has been working to make AI easier to use for enterprises by making the mysteries under the hood more approachable and simpler to understand by users.
And now, armed with $32 million in new Series B funding from investors led by Insight Partners, Fiddler AI is looking to expand and build out its machine learning model performance management platform to bring its services to a broader group of customers.
The new funding from global venture capital and private equity firm Insight Partners, with participation from existing investors, Lightspeed Venture Partners, Lux Capital, Haystack Ventures, Bloomberg Beta, Lockheed Martin and The Alexa Fund, will also be used hire more sales, marketing and engineering team members to continue the evolution of the platform. With this infusion of cash, Fiddler AI has now raised some $47 million to date.
The company’s platform uses explainable AI, which takes the complicated processes behind the scenes and brings in a centralized system to monitor, analyze and explain ML models automatically, Krishna Gade, the CEO of Fiddler AI, told EnterpriseAI. The platform also provides AI transparency and confidence, as well as makes the modeling refinements easier to understand by users, he said.
“AI and machine learning are new technologies that are going to influence business workflows in a big way in many organizations,” said Gade. “The problem with this technology is it is a black box. So, unlike traditional software, where a developer can read the code and debug it or figure out what is going on, AI models are black boxes where you cannot know all the intricacies of how they are making predictions or what is happening behind the scenes.”
That is why the company brings in explainable AI, with specialized toolsets and algorithms that provide insights to customers so they can now see and understand how a particular system is making predictions and how those models can then be changed as needed, said Gade.
“We are giving you this transparency so you can start trusting the system much better,” he said. Customers can use the tools within their own workflows and can then inject their pre-trained models into Fiddler to get needed insights.
Fiddler observes customer models on a continuous basis or can look on an ad hoc basis, he said. The company says its customers include Etsy, BigaBid and several of the largest banking institutions in the U.S. Use cases also include for industrial operations, robotics, credit underwriting, clinical diagnosis, product recommendations, fraud and anti-money laundering and more.
The platform can be used by customers on-premises in their own data centers or as a managed service on the cloud.
The biggest benefit of Fiddler is that it can help customers who are growing their models from dozens to thousands while keeping them manageable and trackable as they scale, said Gade. When large numbers of models are being used, it becomes hard to know and track which models are doing well, which ones are not doing well and when something is wrong, he said.
“This is where explainable AI becomes an enabler,” said Gade. “We provide you with the ability to monitor your models at scale, and to know what is going on with your models, and that is our focus area.”
The Fiddler platform can also help customers with bias problems that can crop up in AI, said Gade. “The models may carry hidden biases because they had been trained on these various datasets. You also need to evaluate fairness and monitor fairness of these models. You have to look at it from a different lens, and that is also something that we help provide. We can ensure that you are monitoring the model on different slices, different segments, such as predicted attributes like race, gender, all kinds of those things to know if the model is performing in a manner that is fair.”
Analyst James Kobielus, the senior research director for data communications and management at research and training consultancy TDWI, told EnterpriseAI that tools for MLOps is a hot field today and that companies like Fiddler AI are adding useful products and services for enterprise customers.
“I find what Fiddler AI is doing to be very interesting,” said Kobielus. “It incorporates a set of promising algorithmic explanation approaches – SHapley Additive exPlanations (SHAP) and Integrated Gradients (IG) – to identify the predictive contribution of various features in a model. If nothing else, this adds a level of mathematical confidence to such explanations that might prove very useful when defending the models for legal and compliance purposes against charges of bias.”
Kobielus called the company “a promising startup, only three-years-old and having gained a fair amount of recognition and capital investment.”
Another analyst, Marc Staimer, president and chief data scientist with Dragon Slayer Consulting, said that Fiddler’s platform expands and accelerates the AI and ML market for customers.
“Anything that makes development and implementation of useful AI and machine learning models into production is a good thing,” said Staimer. “It speeds up development and [the usability] of AL and ML."
For Fiddler AI, success will be measurable if they can really help customers solve their AI problems, said Staimer. “Obviously, their VCs think they have a winner.”