Building a Data Science Platform with Docker

This talk discusses how 3Blades has created a data science collaboration platform using Docker, Docker Compose, and Docker Swarm. This platform allows users to spawn and manage dedicated Jupyter Notebooks, RESTful endpoints and cron services to train, build and deploy machine learning and deep learning models. Docker is a key piece of the platform's technology used to manage this multi-user and multi-team data science environments. 

In addition to demonstrating how these technology pieces work today, the 3Blades team provides context on how they got to the current iteration, as well as provide a roadmap for future docker related endeavors. Specific topics  covered include: deploying docker engine and docker compose to AWS, Dockerfiles to build and manage docker images for the full stack in addition to custom end-user images, use of CI/CD with DockerHub for automated testing and deployment pipelines, among others.