Democratizing data science with Netflix

Human-centric machine learning infrastructure is close to Netflix’s heart. They wanted to contribute to the global field of machine learning by providing an open-source solution for managing data science projects. We had the chance to partner up with them to make it happen.

Towards an accessible data science tool

Metaflow is a tool (a Python library, really) originally developed at Neflix for carrying out data science projects. As we all know, this is at the core of what they do – data science is the magic behind functional algorithms, that is, a regular Netflix user getting better recommendations and finding loads of content they like. 

Before Metaflow was developed, making a data science project happen from scratch required a whole bunch of infrastructure tools and skills, such as cloud architecture, version control, error management and model operations. Metaflow offers a single tool for setting up and managing a data science business project, making it way easier to execute them. 

However, up until now, Metaflow has been a command line based tool, which practically means that using it has required a fair amount of tech knowhow from a casual data scientist. Together with Netflix, we set out to make Metaflow more accessible for data scientists around the globe by creating a graphical user interface for it. 

The perfect equation of freedom and responsibility

From the get-go, Netflix has been the ideal client and co-operation partner. Since day one, they’ve counted on us to get the job done, and trusted us with an ownership of the project. Hence, during the most hectic times of project work, we were in a very good place to handle the unavoidable strain of our team being based in three different time zones: Finland, West Coast USA, and Thailand. Establishing ways for remote working and smooth flow of information early on has proven to be a key factor of success.

Easy to adopt

With the new graphical user interface in place, Metaflow is now much easier to adopt for all data scientists, not just the ones who know the ropes around hard-core tech solutions. The full potential of the tool can now be accessed in a more straightforward way, as the GUI makes the overall user experience much more intuitive. At the same time, there’s way less need for user support.  

A bright future for machine learning experts

Providing a more straightforward user experience for a tool like Metaflow is an act of democratizing data science. The future will show what positive outcomes will spring from organisations and the open source community. Academic endeavours in the field of AI and machine learning will, for sure, become easier for many companies, and the utilization of data will become more efficient. 

“I’d really like to give a big thanks and shoutout to Codemate – they’ve been an amazing partner in developing this. The design work was excellent.”

Brett Rose, Technical Program Manager for ML Infrastructure at Netflix

Want to dig deeper? You can read more about this case from Codemate’s home page