Michelangelo User Guides
Build, train, and deploy machine learning models at scale using Michelangelo's unified ML platform.
Getting Started
New to Michelangelo? Follow this path:
- Getting Started with Pipelines — Build your first pipeline in 30 minutes
- Prepare Your Data — Load, clean, and split datasets using Ray and Spark
- Train & Deploy Models — Train locally or at scale, then register and serve
Quick Navigation
ML Pipelines
| Guide | Description |
|---|---|
| ML Pipelines Overview | Tasks, workflows, pipelines, and the Uniflow framework |
| Running Uniflow Pipelines | Run pipelines locally and remotely |
| Pipeline Running Modes | Local Run, Remote Run, Pipeline Dev Run, and Pipeline Run |
| Project Management | Organize pipelines, models, and resources into projects |
| Pipeline Management | Create, update, and manage pipelines |
| Set Up Triggers | Schedule pipelines on a cron schedule |
| Backfill Pipelines | Reprocess historical data windows |
| Pipeline Notifications | Email and Slack alerts on pipeline run outcomes |
Train & Deploy Models
| Guide | Description |
|---|---|
| Train and Register a Model | Train at scale and register artifacts |
| Model Registry Guide | Version, track, and manage trained models |
| Deploy a Model | Bind a registered model to an inference server |
Reference
| Guide | Description |
|---|---|
| Type System | Uniflow data types and serialization |
| Data Passing and References | How data flows between pipeline tasks |
| CLI Reference | Command-line tools for pipeline and project management |
Examples
| Example | Description |
|---|---|
| California Housing Regression | Predict house prices with XGBoost |
| BERT Text Classification | Classify text with pre-trained transformers |
| GPT Fine-tuning | Fine-tune large language models with LoRA |
| Amazon Books Recommendation | Dual-encoder recommendation system |