Callisto: An ML starter template using Docker

• 

  callisto

Inspiration

Even though the idea of continually training models and deploying models (MLOps) is not new, most of the off-the-shelf solutions come with additional complexities requiring developers/teams to be familiar with various details of a specific cloud vendor. Callisto provides a simple starter kit for training and running ML algorithms using containers that can run locally using Docker.

What it does

Provides a starter template to train and run ML algorithms.

How we built it

Few weekends or stitching together application templates from various platforms such as NestJS, Flask and Angular.

Challenges we ran into

Networking was a challenge as by-design containers do not have access to other containers running in isolation. Using a shared/bridge network helped with connecting the containers that are part of the application.

Accomplishments that we're proud of

Working PoC

What we learned

Docker, LLM with oolama, and networking to mention a few.

What's next for Callisto: An ML starter template using Docker

• [x] Support simple ML training and hosting
• [x] Add Docker support
• [ ] Ollama LLM container support
• [ ] Add GAN model training
• [ ] Add React front-end

Built With

• docker
• javascript
• python