From Hourly Environmental Data to Three-City AQI Forecasts

An end-to-end MLOps forecasting workflow turning hourly AQICN/WAQI and OpenWeather observations into direct 24-, 48-, and 72-hour air-quality forecasts for Karachi, Lahore, and Islamabad.

• Verified Outcome: 105,054 checked-in hourly rows in BigQuery and a strict 606 ordered feature contract between training and FastAPI inference.
• Lifecycle: Chronological train/validation/test evaluation comparing Ridge, Random Forest, and Histogram Gradient Boosting in MLflow, served via FastAPI and Streamlit with scheduled GitHub Actions.
• Designed and developed by Syed Abdullah Bin Masood during an AI/ML and MLOps internship at 10Pearls.