An AI that can lock you out of your own PC every hour or so, and only allow you back in when you did 10 pushups.
It sounded like a good idea before I actually made it.
What better way to learn about MLOps than over-engineering a pet project to a degree where Google can put it into production tomorrow. Using a raspberry pi coupled with a camera / AI accelerator combo kit, I created a fully featured fitness overlord. The full ML pipeline was handled with ClearML, using a Google Pose Estimation model at its core, both open source ofcourse.
The presentation will guide the listener through the journey of creating this overlord, from the specifics on the hardware to the fully automated retraining and deployment pipeline. Mostly the focus will be on the machine learning side of the project, the challenges that I had and how I was able to overcome them using open source tools and industry best practices around managing ML projects.
Speakers: Victor Sonck