Our Scientific R&D portfolio matches (future) technological possibilities and meaningful digital applications. We only conduct scientific R&D when current available solutions are not available or feasible to implement. We currently focus on digital and technology applications. In the future, this may expand to other R&D domains, for example food product development, packaging, or mobility.
Machine learning is about letting computers learn to solve a task by showing them examples rather than explicitly programming them. Machine learning enables the acceleration and improvement of many business processes. Retail has many challenges (transport, forecasting, personalization) that can greatly benefit from novel machine learning solutions. Machine learning allows a search engine to understand the intent of your search queries in an online shop and show you the products that match it best.
In NLP, we teach machines to understand human language. In Dutch, in English, in French, in Czech, in Greek, in any language in which we do business really. NLP matters for Ahold Delhaize because our customers increasingly interact with us by phone, tablet or computer. Our systems need to be smart enough to understand what our customers want and how we can best help them. Understanding our customers is key to our business. And because it is so essential, we need to develop, customize and maintain the required technology ourselves. Entering a search query in an app on your phone should be easy as possible, so that the app understand what you want before you’re doing typing.
This domain is concerned with rapidly trying out novel ideas in a real world context, and measuring their impact. A well-designed experimentation framework allows us to iterate fast, and measure the real world impact of our scientific solutions. Being able to accurately measure phenomena in the real world is at the heart of scientific research. EME allows us to determine whether a new recommendation approach in one of our online shops improves the customer experience, e.g., by allowing customers to find the products that they are looking for faster.
Technology can greatly improve the quality of living of humankind. Yet, history shows that technology should be introduced in a responsible way. For AI-powered products and services this implies they should be safe and unbiased. Also, developing new robotic solutions in a retail context, like cleaning bots, comes with specific ethical challenges. Therefore, we identified this as a key research domain.
Our research domain perception revolves around methods and tools to process sensory signals. This could be image, sound, or actual sensors. As you can imagine, the retail environment offers many opportunities in this area, for example using image recognitions to check the quality of fresh produce
Data engineering is concerned with the collect, storage, processing and validation of large amounts of data. Data engineering provides the necessary infrastructure to manage the data on which modern corporations base their decisions. Bringing machine learning models into production and making sure that they work reliably is an open challenge at the forefront of current research efforts. Data engineering systems will collect and maintain the data from various sources for machine learning applications such as demand forecasting.
Forecasting predicts the amount of sales in the future. Having the right amount of products in stock is a core challenge in retail. An improved forecast does not only help the business, but also the environment for example by reducing food waste. A good forecast makes sure there is enough of your favorite products in stock, even if you come to the store late in the evening!