Whenever change happens, there’s always one department that tries to claim that it won’t work for them. AI teams are the latest to get away with this, hiding behind differences in how research is done to shrug off the requirements of agile development, automated testing and continuous integration. I’ve been working with my team to embed CI even during the initial stages of deep learning and reaped the rewards of improved efficiency for both research and development. It is possible to create a data science development pipeline and I’ll discuss not only how I’ve got this set up, but also how I’ve had to flex the processes to make it work for all involved.
You can see Janet’s slides below: