If you’ve ever searched for a recipe online, odds are you’ve perused at least a few offerings on Yummly. This massive database started out with a focus on personalizing recipe discovery. Then, almost three years ago, appliance giant Whirlpool bought the company and the stakes changed. In the words of Greg Druck, Chief Data Scientist at Yummly, the company has now expanded from “personalizing recipe discovery to personalizing the entire digital kitchen.”
Curious? So were we, which is why we invited Druck to speak at Customize, our food personalization event happening next week in NYC. (Hot tip: There are still a few tickets left, and you can get 15 percent off with code SPOON15.)
To ramp up to the main event we asked Druck a few questions on what exactly a personalized kitchen might look like and what tools it’ll feature (hint: digital assistants and something called a “virtual pantry.”). And yes, the kitchen of the future should be able to perfectly cook a steak to your personal definition of doneness, every single time.
Check out the Q&A below. We’ll see you in New York!
Tell us a little bit about what Yummly does.
Yummly is the most advanced AI-powered digital kitchen platform with over 25 million users. Yummly started out as a personalized platform for discovering online recipes. We are now expanding our offering to support the future of the kitchen. We want to help our users achieve their cooking-related goals with smart appliance integrations, premium guided recipes, and tools for meal planning and shopping.
Yummly places strong emphasis on personalized recommendations for the consumer. How do you optimize those suggestions?
The Yummly recipe ingestion pipeline (pun intended) builds comprehensive representations of over 2 million recipes by inferring latent structure. Machine learning models parse the recipe and map it onto our food knowledge graph, inferring nutrition information, cuisine, techniques, difficulty, and more. This provides a foundation for content-based recommendation algorithms.
Yummly also learns taste profiles for 25 million users by combining explicit and implicit feedback based on behavior and usage. Machine learning systems synthesize this data along with other contextual and ambient signals including day of the week, season, and location to create dynamic personalized feeds for each of our users.
Has Yummly’s acquisition by Whirlpool changed its approach to personalization (by gathering data from home appliance usage, etc)?
Whirlpool’s acquisition of Yummly has allowed us to expand from personalizing recipe discovery to personalizing the entire digital kitchen. We believe personalization is the key to helping people achieve their goals, such as eating healthy, saving money, and reducing stress. Combining Whirlpool appliances as the hardware with Yummly software and machine learning systems allows us to personalize the experience to each home cook.
For example, Yummly will recommend personalized meal plans and shopping lists — in addition to individual recipes — based on a user’s tastes, goals, and appliances. We’ll keep track of the ingredients they have on hand and incorporate information from their “virtual pantry” into recommendations that will help them save money by reducing food waste.
Integrations with the Yummly Smart Thermometer and Whirlpool ovens will allow Yummly to adapt cooking algorithms to each user’s needs: for example, cooking a steak to a user’s personal definition of doneness. Combining these ideas into one seamless experience will substantially reduce friction in the kitchen.
How do you envision recipes (and the recipe recommendation process) getting even more personalized over the next 5 years?
Conversational digital kitchen assistant AIs will help people create plans and recommend custom recipes that are much more personalized to specific needs and more useful for achieving goals in the kitchen. AI will guide you through the week, providing ongoing personalized advice, as well as gamifying and tracking progress against goals over time.
Your AI services will personalize a weekly meal plan and schedule for your household and then have the ingredients delivered to your home. Your plan may include a custom stir-fry recipe that uses up the carrots and chicken that were going bad (recognized using in-kitchen cameras) to save money and reduce food waste. It may avoid pasta or adjust ingredients according to your personalized nutrition plan to help you maintain a low-carb diet (because your assistant knows you’re not tracking well against your weight-loss goal). It might include a cheesy broccoli recipe to help you achieve your goal of getting the kids to eat more vegetables. It may even suggest cooking the chicken dish on Sunday to have an easy meal ready for Monday, reducing the stress of meal planning. Lastly AI may automatically adjust the bake time and temperature to make the dish extra crispy for you, and monitor cooking using a smart thermometer, notifying you when it is done.
This is just the tip of the iceberg — get your tickets to Customize to hear Druck’s fireside chat, where we’ll discuss how personalization will reshape the consumer kitchen. Get 15 percent off tickets with code SPOON15.