Altaviz
MLOps Platform with Medallion Architecture ETL
Tech Stack
PySpark • Delta Lake • PostgreSQL • Streamlit • Plotly • Docker • Terraform
50k+Sensor Readings
10Units Monitored
4Stations
Overview
MLOps platform processing 50k+ sensor readings through medallion architecture with drift detection and retraining.
Problem
Industrial operators rely on reactive maintenance, leading to unplanned downtime. Sensor data across distributed stations is siloed and not actionable for ML models.
Solution
End-to-end MLOps pipeline: ingest sensor data, process through medallion architecture ETL, store multi-window aggregates in PostgreSQL, and monitor model health with automated drift detection.
Architecture
Data Ingestion → PySpark ETL (Bronze/Silver/Gold) → PostgreSQL (7 tables) → Streamlit Dashboard (4 pages)