The problem Jinkies solves

The ecosystem for developing web and mobile applications is very rich with tools existing for assisting developers during every phase of development such as CI/CD and production monitoring. On the contrary, the world of data science/machine learning is still in the dark ages when it comes to best practices of development. The lack of the ecosystem of tools has held back collaboration on ML models.

We present Jinkies, to our best knowledge, the first continuous integration system for machine learning. Building, testing and deployment needs for a ML project is different, Jinkies provides ability to build multiple ML models simultaneously in isolation (using containers), to write complex ML specific tests in few lines and visualize and compare metrics of models with ease. Ultimately, providing better quality assurance and lowering the time needed for performing experimentation and model tuning. Jinkies is built on top of Jenkins, Docker and comes with a new way to write ML tests.

Challenges we ran into

The problem chosen above was so novel and the solution was first of the kind in the world of open source as result of which we had absolutely nothing to take inspiration from and had to develop everything from scartch.