Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: Deployment automation for ML projects of all shapes and sizes (github.com/bodywork-ml)
7 points by alexioannides 7 months ago | hide | past | favorite | 3 comments



I needed a distraction in the Summer of 2020. So, I decided to re-build the MLOps tool that I helped develop for my previous employer - with all the benefits of hindsight factored-in.

Bodywork enables ML engineers to deploy batch jobs and services to Kuberentes, without them needing to know anything about containerisation, Kuberentes and DevOps. Its main aim in life, is to remove the need to build, push and deploy custom Docker images to Kubernetes. Instead, Bodywork configures your cluster to use pre-built Docker images (that we maintain on Docker Hub), that will pull your project’s codebase direct from its remote Git repository (e.g. GitHub), and run it.

All Bodywork users need to do, is develop the executable Python modules that define what each batch job or service needs to do. Then, they add a single config file that describes how to deploy the project, and let Bodywork take care of rolling it out.

Although it was developed with ML pipelines in-mind, Bodywork is capable of deploying almost any executable Python module, as either a job with a well-defined end, or as a service with no scheduled end (a long-running process). It’s basically a Python wrapper around Kubernetes primitives and we’re hoping that it could also find a place in the scientific Python and quantitative finance communities.


This works really well, a simply effective tool.


Thanks - simplicity and focus is our aim!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: