March 12, 2018
Case Study: How Differential Privacy Became an Integral Part of Integrate.ai’s Product Vision
When integrate.ai needed a way to put privacy guarantees in place for its new AI platform, it turned to the Georgian R&D team for help.
Toronto-based integrate.ai has an ambitious agenda. The company is building an AI-powered platform for B2C enterprises that’s designed to integrate with business processes to make consumer interactions more natural and valuable. To achieve that, its platform brings data together from multiple sources to provide richer insights into how consumers behave in different business settings.
Of course, integrating data from different environments presents new privacy challenges. That’s because traditional de-identification techniques designed to protect personally identifiable information don’t provide strong guarantees since it’s possible to cobble together an individual’s identity using a variety of different sources. To avoid this issue, integrate.ai knew that it could use differential privacy, a mathematical definition for the privacy loss that results to individuals when their private information is used to create a data product. Differential privacy not only offers a solution to potential privacy issues, it’s also an approach that’s well-suited to machine learning products.
Unlike general purpose software development, implementing differential privacy can be a long process that requires considerable applied research expertise to get right. The integrate.ai team needed to find a way to speed up that process to bring its product to market with very clear data privacy guarantees in place. And, it needed to find a way to do so without tying up its R&D resources.
Making an Impact
After making a $5 million seed investment in integrate.ai in early 2017, Georgian started working with the company to accelerate its business. Soon after, the Georgian R&D team began working with integrate.ai to create the differentially private AI solutions it needed. By adopting Georgian’s differential privacy product, integrate.ai was able to demonstrate that it could build a differentially private machine learning model for one of its customers while maintaining similar performance.
The result was that integrate.ai was able to add differential privacy to the machine learning model it had built for Kanetix Ltd., Canada’s largest digital customer acquisition platform for insurance and financial services. That model had been designed to help Kanetix target consumers on its car insurance comparison website who were undecided about making a purchase. Using the website’s deep pool of data, integrate.ai was able to configure its platform to predict which customers were most likely to convert and what their preferred method of transaction was. This enabled Kanetix to begin tailoring the buying experiences it created for those customers. While the differentially private version of the model never went into production, it helped validate that the company could get similar results to those it had already been getting, but with the added benefit of differential privacy.
“The Georgian team was a huge help to our business,” says Sandy Ward, Technical Team Lead at integrate.ai. “Thanks to their product, we’re now able to build differential privacy into our overall product vision and architecture with no real risk to our business. That gives us the ability to create incredible insights for our customers that we wouldn’t have otherwise been able to create.”
Today the integrate.ai team is also much more knowledgeable about differential privacy than it was just a year ago. Knowledge transfer has been an important additional benefit the company derived from working with the Georgian team.
“It’s great to see another one of our portfolio companies adopting our differential privacy product,” says Chang Liu, Applied Researcher. “We’re delighted to see our product providing significant results and helping companies differentiate themselves.” For integrate.ai, it’s all part of a commitment to privacy and trust. The company’s strategy is to invest heavily in the latest techniques to help enterprises succeed with AI in a way that’s focused on reducing risks, rather than having it be an after thought.
So far, it’s a strategy that has really paid off.
“The Georgian R&D team was a huge help to our business. Thanks to their product, we’re now able to build differential privacy into overall product vision and architecture with no real risk to our business. That gives us the ability to create incredible insights for our customers that we wouldn’t have otherwise been able to create.”
Technical Team Lead, integrate.ai
in your inbox
Join our community of thousands of tech entrepreneurs to get actionable insights from our monthly newsletter.