A data engineering platform for live weather and air-quality intelligence on any city.
- role
- Solo build
- status
- Shipped
- when
- 2026
- ownership
- Solo
Who did what. Personal project.
Overview
Atmos ingests live weather and air-quality data for cities worldwide, validates and transforms it through an ETL pipeline, stores it relationally, scores it, monitors its own runs, and shows the results on a dashboard.
Why it exists
Most data projects stop at a notebook. Atmos was built around the parts real data systems need: validation, monitoring, reproducible deployment and failure tracking.
My part
- Built extract, validate, transform and load stages with Pydantic schemas and SQLAlchemy models.
- Built weather, AQI and readiness scoring with risk classification.
- Tracked every pipeline run (runtime, success, failure) and exposed it in a Streamlit dashboard.
- Containerised with Docker and Docker Compose; Typer CLI.
How it works
- a
Extract
Geocoding, weather and air-quality APIs.
- b
Validate
Pydantic schemas and data-quality checks.
- c
Load
SQLite via SQLAlchemy: cities, weather, AQ, metrics, runs.
- d
Analyse
Readiness and risk scoring per city.
- e
Observe
Run monitoring, CSV exports and dashboard.
Built with
- Python
- Pandas
- Pydantic
- SQLAlchemy
- SQLite
- Streamlit
- Typer
- Docker
Status & next
Historical trend views.
Source: github.com/AA-KH/Atmos