Last weekend I participated at the Open Food Hackdays in Basel. As on any hackathon I have joined so far, it meant a lot of fun, new acquaintances and most of all, fast-track learning. This time, I learned some valuable lessons on machine learning.
I joined the team prognolite, which has already been existing as a startup and was already very successful at the previous Open Food Hackdays in Lausanne. Enriched with a few data scientists, a UX specialist and me in the role of frontend dev, we hacked away for 36 hours. The goal was a mobile app that helps restaurants forecast the number of customers for the next few days, based on historical data of the same restaurant. This ultimately lead to a functional app prototype called Antonios Kitchen Solution.
While I already have been peeking into data science, this project was really a revelation to me. Mostly because of the tangible, practical way we employed machine learning, in a much more focused scope than big data or business intelligence. That was new to me. Not just doing a one-off analysis, but something repeatable, well encapsuled. This is music to a developer’sears. Data scientists create (or rather train) ML models, and we make them available as services, just like a library.
Machine learning is not just about big data or business intelligence. It can be employed at a very practical level, e.g as microservices being consumed by traditional transactional applications.
This impression of ML being here and now, and way past its ‘academic’ phase has been energizing me since the Hackdays, and this article ist just the preface to a series of blog articles to come (and hopefully something on GitHub as well).
Just as a hint on what might come: Serving Scikit-Learn ML models as a REST API with a Flask server seems to be fairly straightforward (and without dependencies to cloud providers). We already implemented this architecture approach in part during the hack, but abandoned it in favor of an API that was already in place.
Thanks to the whole team and the OpenData guys for the amazing experience, thanks to Jutta for being in the photo below with me, and to my employer IWF Web Solutions for letting me go to the hackdays partly during working time!
Photo: Jutta Jerlich from BaselHack and me