Paulescu/hands-on-train-and-deploy-ml
A hands-on MLOps course disguised as a crypto bot
This repo teaches the full machine-learning lifecycle—from training to automated deployment—using crypto price prediction as the motivating example.
Not currently ranked — collecting fresh signals.
star history
What it does
This repository is a hands-on tutorial that walks you through training a machine-learning model on historical crypto data, deploying the best experiment as a serverless REST API, and automating the pipeline with GitHub Actions. It is split into three lectures: model training with CometML experiment tracking, deployment via Cerebrium, and continuous deployment triggered by the CometML Model Registry. The author explicitly warns that the goal is mastering MLOps tooling, not building a profitable trading system.
The interesting bit
The project uses the notoriously difficult problem of crypto prediction as a teaching sandbox, keeping the tutorial engaging while staying honest that the resulting model will not make you rich. It also stitches together the entire lifecycle—from raw data to registry-triggered serverless deployments—into a single narrative rather than the usual fragmented snippets.
Key highlights
- End-to-end flow covering data fetching, feature engineering, training, REST API deployment, and CI/CD automation
- CometML serves as both the experiment tracker and the model registry, with webhooks that kick off deployments
- Cerebrium provides the serverless hosting, so the tutorial avoids infrastructure setup
- A Makefile-driven quickstart attempts to compress the setup into a few commands
- Framed as an educational MLOps walkthrough, not a reusable trading library
Caveats
- The README is largely a navigation shell; the detailed instructions live in three separate lecture markdown files
- The “5 minutes” quickstart actually lists seven steps, requires API keys for two external services, and manual copy-pasting of a deployment URL
- Tight coupling to CometML and Cerebrium means the workflow is less portable to other MLOps stacks
Verdict
Good for data scientists and ML engineers who want a guided path from notebook experimentation to automated serverless deployment. Not useful if you are looking for a standalone crypto-prediction library or an open-source-only stack.
Frequently asked
- What is Paulescu/hands-on-train-and-deploy-ml?
- This repo teaches the full machine-learning lifecycle—from training to automated deployment—using crypto price prediction as the motivating example.
- Is hands-on-train-and-deploy-ml open source?
- Yes — Paulescu/hands-on-train-and-deploy-ml is open source, released under the MIT license.
- What language is hands-on-train-and-deploy-ml written in?
- Paulescu/hands-on-train-and-deploy-ml is primarily written in Python.
- How popular is hands-on-train-and-deploy-ml?
- Paulescu/hands-on-train-and-deploy-ml has 888 stars on GitHub.
- Where can I find hands-on-train-and-deploy-ml?
- Paulescu/hands-on-train-and-deploy-ml is on GitHub at https://github.com/Paulescu/hands-on-train-and-deploy-ml.