Machine learning helps Orbisk combat food waste
Orbisk was founded in February 2019. To fight food waste, they developed prototypes with hardware and software based on deep learning and automated image recognition to identify and measure food waste in (large) kitchens of corporate caterers, and hotels. In order to continuously improve and automate their software and data in the most efficient and future-proof way, machine learning workflows were needed.
Orbisk’s key challenge was: “How can you operationalize ML workflows to deploy ML models as efficiently and scalable as possible?”
Orbisk takes part in the initiative Google Cloud for Startups. Being a satisfied user of the Google Workspace applications, choosing Google Cloud and its Premier partner g-company was a logical step to explore the possibilities of machine learning within Google’s cloud.
Working with Orbisk, g-company shares a lot of expertise and methodology. Johanna Schacht, Team Lead Artificial Intelligence, states: “During one of the sessions, we asked questions back and forth for eight hours. That was super valuable.”
The final solution is the automation of the entire ML infrastructure, and the quality control of both the data and the model. This infrastructure is designed with the open-source platform Kubeflow and Tensorflow Extended, hosted on Google Cloud Platform.
Due to COVID-19, all Orbisk’s customers have been closed in spring 2020 and are still closed as we speak (January 2021). Therefore, the automation of their infrastructure and of the data and model quality control will run - for the time being - according to the prototype. The full development and deployment of the ML infrastructure will be postponed until after the corona crisis.
Despite the impact of corona, Orbisk continues their research on further innovating the software. They are optimistic: if the catering industry can open its doors again within a couple of weeks or months, they will be fully engaged in receiving all the data and proactively responding to it.
Their goal is to develop customer-specific models in the future, such as for a bakery. This type of customer buys many different food products, other than a corporate caterer.
Orbisk is happy to tell you more about this success
Reach out to us and we will be happy to get you in touch
with this customer.
Meet the Data Analytics Squad
Thomas van Latum
Curious to know how we can help your organization?
Contact our account manager Jeroen