IceCream Labs uses machine learning and visual intelligence to drive ecommerce

There are no over-the-top flavors or crunchy mix-ins being tested at IceCream Labs, no dream team tasked with creating the next frozen concoction. Still, Madhu Konety, founder and CEO of the artificial intelligence startup headquartered in San Francisco, considers the solutions developed by IceCream Labs to be both sticky and sweet.

In just over two years, the company has leveraged the power of machine learning and visual intelligence to improve content, increase consumer engagement, improve merchandising productivity and accelerate revenues at two of the world’s biggest retailers.

“We enable retailers to do intelligent merchandising around ecommerce,” Konety says. “In simplest terms, we look at product data, marketing data and consumer data and help retailers and brands target the right products to customers. Our solutions also help users find exactly what they’re looking for when they’re shopping online.”

IceCream Labs uses the subset of AI called neural networks, or deep learning, to build these models. Konety says deep learning networks can recognize patterns in large amounts of data and continue to provide cognitive visibility as more information is amassed and more queries are made.

“The applications on our platform continuously process and profile the content quality of over 100 million products. There are also 50 million images empowered by our big data algorithms. This is how the neural network works.”

Madhu Konety, IceCream Labs

“We’ve developed algorithms that teach machines to learn like humans,” he says. “If you were teaching a child what a chair was, you would have to show them several different types of chairs. Eventually they will learn the attributes of a chair and recognize it in various forms.

“That’s what’s happening, but obviously the scale in the case of a retailer is much larger,” Konety says. “The applications on our platform continuously process and profile the content quality of over 100 million products. There are also 50 million images empowered by our big data algorithms. This is how the neural network works. It’s essentially mapping human neurons onto machines.”

IceCream Labs’ catalog management, category management and consumer insights systems deliver on-demand intelligent merchandising for retailers, brands, suppliers and manufacturers. Specifically, the catalog management piece uses machine learning models to automatically enhance content listings, while category function fills in product gaps with market intelligence. The consumer insights tool allows users to assess impactful customer insights capable of driving revenue.

“Currently our focus is really around the merchandising space,” Konety says. “The two big things that we solve are helping retailers understand what consumers are looking for and making sure that they’re presenting the information and product that is relevant to that consumer.”

IceCream Labs systems raise the bar on relevancy by making sure that the product descriptions are relevant for diverse shopper demographics. “We can create descriptions in multiple writing styles, so a retailer can have a millennial style, or a Gen X quality,” Konety says. “It’s not just about personalizing the product assortment but also personalizing the product content. These are the kind of things that the neural net can do.”

Though relatively new to the retail scene, IceCream Labs started big — working with a “very large” retailer looking to increase product assortment by a multiple of 100. Konety reports the programs it put in place have delivered a significant bump in revenue. By making sure that consumers are seeing complete product information, the retailer has achieved from 20 percent to 200 percent increases in revenue — depending on the product.

Today, most of the startup’s customers are big-box retailers, though specialty retailers are awakening to the potential of AI and machine learning technology. Konety says what differentiates IceCream Labs is not only its ability to work with large data sets but speed of implementation and the fact that it runs in the cloud — eliminating software installs and security hiccups.

Konety says the startup has a team of computer science geeks with deep experience building complex enterprise solutions — and very little hair. “We eat and breathe AI and that type of talent is not easy to come by,” he says. Moreover, it allows IceCream Labs to amortize the cost over multiple customers.

Konety’s objective for 2018: Broaden the product offering allowing the company to solve more problems for existing customers and add new ones to the mix. In 2017, IceCream Labs achieved 100 percent year-over-year growth. This year, Konety and his team have set their sights on 300 to 400 percent growth.