Learn how Wellio leverages the Appen platform to power the first intelligent platform that helps people plan, shop, prepare and enjoy healthy meals at home.
Wellio is the first intelligent platform that helps people plan, shop, prepare and enjoy healthy meals at home to support people eating better. Wellio has chosen to help people cook their own meals because individuals control what goes into their food, and they enjoy the actual cooking practice. By providing customers with the expertise of a personal chef and nutritionist and algorithmically curated recipes, Wellio makes it more convenient to eat at home and achieve nutrition goals.
To achieve their goal of helping people cook at home, Erik Andrejko, Wellio’s CTO said that the company “is building a platform that embodies the intersection of culinary and health expertise.” Doing so required building machine learning algorithms that have the same capacity an expert does to make inferences and suggestions about nutrition and cooking. Building machines that act like experts involves vast volumes of data in the form of things like images of food and recipe information.
However, just having mountains of information is not enough. Machine learning algorithms need annotated training data. Wellio was, by its Andrejko’s own admission, “data-rich but label poor.” The company had collected large quantities of unstructured data, such as recipes and images of different types of food. Data scientists were able to develop algorithms with some accuracy without unlabeled data. But, to mimic the expertise of human nutritionists and chefs, the algorithms needed a higher degree of accuracy — accuracy that wasn’t achievable without labeled data.
In order to generate the accuracy needed to ship its product, Wellio understood it needed a solution to label data at scale. It became clear that we offered a good solution.
Wellio leveraged the Appen platform for data categorization and text annotation, in order to turn their raw data sets into labeled training data.
Wellio is, as its VP of Data Science, Sivan Aldor-Noiman pointed out, now “able to not only perform one-time data annotation at scale but is also able to create data pipelines where new predictions and trainings are continuously flowing.”
Part of this pipeline included leveraging actual chefs for an “experts-in-the-loop” process, whereby chefs participated in the data annotation process.
Wellio feels it now has a true strategic advantage: It can weave data annotation into its normal engineering pipelines and focus on growing its business.