Tech Stack

Tech Stack

MLOps pipeline for air quality prediction.

Table of status of each step

Step	Tool	Status
Package Manager	uv	✅ Done
Data Storage	CSV → Parquet	🔶 CSV only
Data Processing	Pandas	✅ Done
Data Versioning	DVC	✅ Done
Data Validation	Great Expectations	❌ Not started
Feature Store	Local Parquet	❌ Not started
Experiment Tracking	MLflow	✅ Done
Orchestration	Python scripts	✅ Done
Model Training	Scikit-learn + XGBoost + LightGBM	✅ Done
Hyperparameter Tuning	Optuna	❌ Not started
Model Registry	MLflow	✅ Done
Model Serving	FastAPI + Docker	✅ Done
Monitoring	Evidently	✅ Done
Cloud Deploy	Railway	✅ Done

Table for comparison of local vs production tech stack for MLOps

Step	Local or Free Tool	Production Equivalent
Package Manager	uv	Same
Data Storage	CSV → Parquet	S3 + Delta Lake
Data Processing	Pandas	Apache Spark
Data Versioning	DVC	Delta Lake / LakeFS
Data Validation	Great Expectations	Same
Feature Store	Local Parquet	Feast / Tecton
Experiment Tracking	MLflow	Same / W&B / Neptune
Orchestration	Python scripts	Apache Airflow
Model Training	Scikit-learn + XGBoost + LightGBM	Same + Ray / Dask
Hyperparameter Tuning	Optuna	Same
Model Registry	MLflow	SageMaker / Kubeflow
Model Serving	FastAPI + Docker	KServe / Seldon
Monitoring	Evidently	Arize / Fiddler
Cloud Deploy	Render / Railway	AWS / GCP / Azure