We took a deep dive into various use cases around machine learning, data science, and artificial intelligence at Wolt. From logistics optimization to personalization to optimizing marketing data platforms, there’s a world of interesting data science challenges in building a local commerce platform that has been uncovered at this meetup.
At the start of the meetup, Dirk Hofmann, Co-founder of DAIN Studios, welcomed the Wolt team onto the stage. Erlin Gulbenkoglu, a Data Scientist at Wolt, acted as moderator for the summit which included talks by Aatu Kaapro, a Data Science Lead at Wolt, Juan Orduz a mathematician and data scientist who forms part of the Wolt marketing team, and Aleksandr Fedintsev, a staff data scientist. There was a lot to learn about day-to-day Wolt operations.
Three insiders discuss data impact at Wolt
Aatu Kaapro started proceedings by explaining what Wolt does with data science. The talk then gave an overview of the firm’s transition from delivering just restaurant food to delivering other items such as retail products. Aatu stopped to mention how the increase in the volume of users and the range of services provided by Wolt has made the company’s work more challenging.
Aatu Kaapro is a Data Science Lead at Wolt. With more than 10 years experience delivering data science and machine learning solutions, Aatu leads the Wolt data science practice.
After Aatu had finished, Erlin asked Juan Orduz to take the mic. He talked about how measurement is central to the optimization of marketing efficiency. He focussed on the importance of good experimentation, simulations, and learning—comparing and contrasting the advantages of a number of different approaches. There were some interesting questions from the audience including one about getting hold of and using a control group. This drew an insightful response from Juan who talked about the principle of unknown unknowns.
Juan Orduz is a mathematician and data scientist who forms part of Wolt’s marketing team. He leads data science projects from conceptualisation and modeling to deployment.
During the final session, Aleksandr Fedintsev highlighted the importance of personalization at Wolt. He also provided insights on offline metrics, machine learning, and neural networks. Towards the end of his speech, he offered an unexpected solution to latency spikes. (Watch the recording to believe it.)
Aleksandr is a data scientist and machine learning engineer with a passion for statistics, recommender systems, deep learning and low latency online inference. At Wolt, Aleksandr works as a staff data scientist in the Consumer Group helping several teams to make customer experience across multiple touchpoints more personalized.
The agenda (EET)
Inspiration. Learning. Networking.
Taken as a whole, the HDSM and Wolt gathering was a great success. There was plenty of inspiration and learning and development on offer. Attendees—online and in person—got the chance to chat with industry experts and delve into complex real-world use-case scenarios.
During the post-talk mingling session there was also the opportunity to make friends, share experiences, and grow and diversify contact networks.
If you’re interested in hosting a meetup, please contact us and let’s plan the next edition of the Helsinki Data Science Meetup together. Alternatively, if you want to learn more about the collaboration between Wolt and HDSM, you can rewatch the event here or on our YouTube channel.