Halla taps AI for personalized grocery and dish recommendations | Tech Industry

You just got home after a long commute from work, and you’re starving. The pantry’s empty — you didn’t get around to buying groceries — so you fire up a food delivery app and hop over to a list of favorites. You’re in the mood for something different, but you’re stuck with indecision; what if you order something you end up disliking?

Los Angeles-based startup Halla aims to solve that problem once and for all with Halla I/O (which stands for “intelligent ordering”), a platform that uses artificial intelligence (AI) to generate Netflix-like recommendations for grocery, restaurant, and food delivery apps and websites.

“We use psychographics and data to predict your preferences and cravings at any given moment,” CEO and cofounder Spencer Price told VentureBeat in an interview, “and we apply predictive analytics and AI to taste and flavor attributes to gain an understanding of the food itself.”

To do that, Halla’s algorithms tap a proprietary database of more than 10,000 grocery items, 20,000 ingredients, 175,000 recipes, and 20 million restaurant dishes. They take into account the ingredients (and even the molecular makeup) of dishes and recipes, building a taxonomic map of attributes like flavor, appearance, and dietary appeal.

Here’s a concrete example: Say you love chicken sandwiches. Basic machine learning models might infer that you aren’t opposed to, say, chicken piccata, but Halla I/O considers the bigger picture. It recognizes that your preference for a chicken sandwich likely isn’t just about the type of protein, but about the bread, the toppings, and the experience of eating it from your hand. Therefore, it concludes that you probably like subs and burgers, too.

“Food is very psychological. Every dish breaks down into tremendously data-rich subcomponents, but then deciding which elements matter to a person selecting a dish or planning a home-cooked meal requires a logic particular to food,” Price said. “When you focus only on data science without psychology, you miss out on the most fundamentally valuable aspects of food decisioning.”

Halla I/O adopts a software-as-a-service model — customers integrate its recommendation engine through an API that can be accessed through a subscription. The company is piloting its service with four brick-and-mortar stores, including Green Zebra, a grocer in Portland, Oregon that uses it to drive in-store display decisions. If a particular brand of jam started flying off the shelves, for example, Halla I/O’s system would recommend stocking bagels and bread on a nearby shelf.

Other launch partners are tapping Halla’s contextual smarts to improve their in-app shopping experiences. Because Halla I/O retains anonymized purchase data, it can automatically remove items you’ve recently bought from its recommendations. And thanks to its database of epicurean knowledge, it’ll recognize when you’re shopping for a recipe and recommend relevant ingredients.

“One of the questions we’re trying to answer is how do you make it so that the in-store experience is as efficient and personalized as possible?” Price said. “The average grocery store sells 34,000 products, but the average customer buys just 300. The goal, then, is to tailor each section to specific customer segments.”

There’s more to this than just a backend. Halla’s app for iOS leverages natural language processing and partnerships with Delivery.com and EatStreet to recommend dishes from 25,000 restaurants across the country.

Whether the multipronged approach is enough to give Halla a leg up on other AI-forward food recommendation startups, like Dishq, Plant Jammer, and Foodpairing, remains to be seen, but Price is confident that the core technology — and Halla’s commitment to customers — sets it apart from the competition.

“At the end of the day, it’s about narrowing the selection from hundreds of choices to the ones that make sense for you,” Price said.

Halla has just seven employees and has raised $500,000 in funding, to date.