AQPearls AQIIslamabad Forecast LabLaunch dashboard
Salman Khan

Developed by

Salman Khan

AI Engineer & Full Stack DeveloperFocused on turning machine-learning pipelines into scalable, production-grade, and decision-ready software applications.

LinkedInGitHubLinktree

Project Overview

Pearls AQI Predictor

Pearls AQI is a production-grade air quality forecasting system for Islamabad. It acts as an automated, self-healing pipeline that runs continuously. Rather than relying on static datasets, it collects live meteorological and pollutant readings, trains four machine-learning models daily, registers the top-performing candidate, and serves real-time predictions via a robust API.

Inference Latency< 85ms avg
Ingestion CycleHourly (24/7)
Retraining CadenceDailyRetrain
Model RepositoryMongoDB GridFS
Islamabad AQI Forecast Lab

System Strengths

Pipeline Engineering

  • Hourly live feature collection with duplicate guards and quality audits.
  • Daily automated retraining and scoring of four distinct ML algorithms.
  • Secure model artifact registration and versioning in MongoDB GridFS.
  • Time-series validation split to eliminate data leakage and guarantee temporal honesty.

Stack

Technologies Used

PythonScikit-LearnFastAPIMongoDB AtlasGridFSGitHub ActionsNext.jsTypeScriptTailwindCSSRecharts

Ready to inspect

Explore predictions, check model performance, or analyze data quality live.

Launch Dashboard