The sweetness world knows him as one among the pioneers who drove augmented reality for the sector at Modiface. Now Parham Aarabi’s next act also looks like a technical beaut, and never only for makeup. The tech start-up founder and University of Toronto professor of artificial intelligence goes full bore into e-commerce along with his latest enterprise, Predictive Commerce.

With the platform, which launches Tuesday, Aarabi does what he does best — determine how AI can solve real-world problems. Here, the corporate predicts human behavior, so brands can adapt their web sites to realize maximum effect.

The genesis of the business didn’t stem from a selected eureka moment, but a simmering curiosity that nagged at him, he told WWD. He kept noticing how some retailers and types struggle with the nuances of e-commerce and others didn’t. He just needed to know why.

“It had bugged me for years, back at Modiface and particularly at Modiface through the L’Oréal years that I stayed on as CEO [chief executive officer], that it was at all times a mystery,” he said, referring to L’Oréal’s acquisition of the wonder tech firm in 2018. “Same tool, similar sites, but different results.”

After he left Modiface in 2022, he had loads of time to satisfy his curiosity. As professor of AI, he also had access to a team that would help solve the mystery.

“That team laser-focused on this key problem of simulating sites, and in doing these simulations, we were in a position to answer questions as to why, in certain cases, a virtual trial tool may also help and why in other cases it actually may or may not help,” he continued. The particulars can change quickly, depending on any variety of things, like market forces, trends and other outside aspects. But he’s been constructing an AI that may sustain.

What helps is that others are only as curious, even vexed, concerning the same scenario.

The firm got here together in earnest last January, and it has been constructing the core technology. But just three months later, while it was still in stealth mode, it someway racked up nearly 80 clients. A large portion of the interest might be resulting from the firm’s predictive powers.

In keeping with Aarabi, Predictive’s accuracy rate was 50 percent in January, and it’s at a surprising 94 percent now. He expects to achieve 99 percent by the summer. Predictive claims that its AI-driven platform is 1,200 times faster and 100 times cheaper.

The key sauce is a mixture of information scraped from the net, including underlying information, images and other facets, so it could actually feed the model, plus reciprocal data from the brands it really works with. If a client also comes back with a human testing result, Aarabi compares that to his models to enhance the algorithms and the accuracy rate. With regards to AI, especially machine learning, the more data the higher.

Essentially the most obvious utility for the tech is A/B testing, which is a very important a part of maintaining any website or online experiences — and yet, it’s an area that’s often neglected.

“Most brands don’t A/B test enough; they go based on gut feelings, because A/B testing is dear,” he contended. “[Our simulations] allow them to try so many various ideas and optimize their site profoundly greater than what it’s today. We’re seeing an amazing impact in the standard of web sites based on the start line and after a number of rounds of optimization.”

People might think beauty brands that already know Aarabi from his 15-plus years at Modiface, before he left in September, can be his important priority. But they might be mistaken.

“Surprisingly, it’s not the wonder brands, but fashion is one among our important goals. I feel our largest vertical straight away is the style space,” he added. “The second to that might be just non-beauty, non-fashion CPG [consumer packaged goods] brands which have sites, after which optimizing them is sort of key.”

Several of his clients are Fortune 500 brands. Based on his experience, there may very well be an excellent reason for that.

“They wish to do it very routinely, so each time they make a change on the location, they arrive to us and see if that’s good or bad.”