How danielfelromero/aa-dev works
AA Dev is a teaching example of an analytics Lakehouse and churn-prediction workflow, not a production platform. Its main path ingests a CSV into Bronze, transforms and pseudonymizes it into Silver, prepares Gold features, trains a model, and serves predictions; a separate observation script checks data quality and drift. The graph follows the sampled script calls and file reads/writes. README-described capabilities such as service observability and broader Lakehouse trade-offs are not expanded into unsampled implementation details.
Main entry points: 02_transform.py Silver Transform · 01_ingest.py Bronze Ingestion · 03_train.py Model Training · 05_observe.py Data Observation
Components 14
Lakehouse Data
- Bronze Ingestion
scripts/01_ingest.py - Bronze Delta
scripts/01_ingest.py - Silver Transform
scripts/02_transform.py - Quality and Privacy
src/aa_dev/quality.py - Silver Delta
scripts/02_transform.py - Feature Dataset
src/aa_dev/features.py - Gold Parquet
scripts/02_transform.py
ML Workflow
- Model Training
scripts/03_train.py - Trained Model
scripts/03_train.py - Prediction API
scripts/04_serve.py
Observation
- Data Observation
scripts/05_observe.py - Quality and Drift
src/aa_dev/monitoring.py
Other components
- Customer CSV
data/source/customers.csv - API Client
Connections 13
- Customer CSV → Bronze Ingestion: reads CSV
scripts/01_ingest.py - Bronze Ingestion → Bronze Delta: writes Delta
scripts/01_ingest.py - Bronze Delta → Silver Transform: reads Delta
scripts/02_transform.py - Silver Transform → Quality and Privacy: calls transform
scripts/02_transform.py - Silver Transform → Silver Delta: writes Delta
scripts/02_transform.py - Silver Transform → Feature Dataset: builds features
scripts/02_transform.py - Silver Transform → Gold Parquet: writes Parquet
scripts/02_transform.py - Gold Parquet → Model Training: reads dataset
scripts/03_train.py - Model Training → Trained Model: saves model
scripts/03_train.py - Trained Model → Prediction API: loads model
scripts/04_serve.py - API Client → Prediction API: requests predictions
README.md - Gold Parquet → Data Observation: reads dataset
scripts/05_observe.py - Data Observation → Quality and Drift: calls metrics
scripts/05_observe.py