Real-life Machine Learning systems: fashion retail & improving a ML workflow

We introduce a couple of cutting-edge Machine Learning talks in this meetup! Come learn how to train your model building real-life machine learning systems!

Event schedule:

18:00 Doors open
18:15 Pizza and beer 🍕🍻(Big thank you to de Bijenkorf for hosting us!)
18:45 Intro by Codemotion
19:00 Talk #1: Outfit Recommendations: How to use existing data/images to create new outfits, by Jelle Rolf (Data analyst e-commerce at de Bijenkorf https://www.linkedin.com/in/jelle-rolf-949156b0/) and Dennis van der Voorn (Data analyst at de Bijenkorf https://www.linkedin.com/in/dennis-van-der-voorn-6730bb94/)
19:45 Talk #2: Improving Machine Learning Workflow - Training, Packaging and Serving your Models, by Wilder Rodrigues (Artificial Intelligence Engineer at Aigent https://www.linkedin.com/in/wilderrodrigues/)
20:30 Drinks and networking

➡️Talk #1: Outfit Recommendations: How to use existing data/images to create new outfits

Gaining useful insights into unstructured data, and specifically images, is becoming increasingly important in a large number of markets, including E-commerce. At the same time the field of computer vision is an extremely vast area with a steep learning curve to actually become proficient in. In this talk we will elaborate on how we use image similarity at de Bijenkorf.
First, we will take you through the steps we took to set up an Image Similarity algorithm that is useful for fashion retail. We take you through some methods we tried and present our findings about these methods.
In the second part we will show you how we use images of outfits that are already available, to generate new combinations of products. We will go through how to prepare the data from the images, the methods we used and the advantages & disadvantages of the steps we took.

➡️Talk #2: Improving Machine Learning Workflow - Training, Packaging and Serving your Models

As machine learning practitioners, we know how hard it can be to have a smooth process around training and serving production-ready models. Processing the data, saving all the relevant artefacts to make experiments reproducible, packaging and serving the models; all these individual components can be a nightmare to implement and manage. MLflow - an amazing new platform for managing the ML life cycle - comes to the rescue.

In this talk, we will present a Docker powered infrastructure that combines MLflow, JupyterHub and Minio (S3 compliant storage) that aims to solve the above problems and improve your machine learning workflow. In addition, we will present a CI-CD pipeline which is responsible for fetching production-ready models from storage, and building and publishing Docker images that serve these models in production. With this in place, tasks like experimenting, releasing and serving models become more straightforward and less manual. We will explore how this infrastructure can speed up our work, make it less error prone, and help us manage all ML related artefacts better.

We will start the talk by presenting the infrastructure and its components and how they address practitioners’ pain points. Next, we will show how our solution helps to train models in a structured way. And lastly, we will demonstrate how to automate packaging and serving of the models prior to deployment.

Thanks to De Bijenkorf for having us!