🩺 Kidney Disease Classification – MLflow & DVC

An end-to-end kidney disease classification system with experiment tracking, reproducible pipelines, and production-ready ML workflows.

🌐 Live App • 🔁 Workflow • 🏗️ Architecture • 🚀 Quick Start • 📊 MLflow & DVC

---

🎯 Overview

This project is a production-style machine learning system for kidney disease classification, designed with MLOps best practices in mind.

It focuses on:

• Modular ML pipelines
• Experiment tracking with MLflow
• Reproducibility using DVC
• Configuration-driven development
• Deployment-ready Streamlit interface

The project structure closely resembles real-world industry ML systems.

---

🔁 Workflows

The entire pipeline is configuration-driven and modular:

1. Update config.yaml
2. Update secrets.yaml (optional)
3. Update params.yaml
4. Define entities
5. Update configuration manager in src/config
6. Update individual components
7. Update pipeline logic
8. Update main.py
9. Update dvc.yaml
10. Update app.py

This design ensures clean separation of concerns and easy experimentation.

---

🏗️ Architecture

Data Ingestion
      ↓
Data Validation
      ↓
Data Transformation
      ↓
Base Model Preparation (VGG16)
      ↓
Model Training
      ↓
Model Evaluation
      ↓
MLflow Logging & Registry
      ↓
Streamlit Inference App

DVC orchestrates the pipeline, while MLflow tracks experiments and models.

---

🛠️ Tech Stack

Layer	Technology
Language	Python 3.10
Model	VGG16 (Transfer Learning)
ML Framework	TensorFlow / Keras
Experiment Tracking	MLflow
Pipeline Orchestration	DVC
Frontend	Streamlit
Configuration	YAML
Tracking Server	DagsHub

---

🚀 Quick Start

Prerequisites

• Conda
• Python 3.10
• Git
• DVC

---

Clone the Repository

git clone https://github.com/vivek34561/kidney_disease_classification
cd kidney_disease_classification

---

Create Conda Environment

conda create -n myenv python=3.10 -y
conda activate myenv

---

Install Dependencies

pip install -r requirements.txt

---

Run the Application

python app.py

Open your browser and navigate to the local Streamlit URL shown in the terminal.

---

📊 MLflow & DVC

MLflow

• Tracks experiments, parameters, metrics, and artifacts
• Maintains model registry
• Enables reproducibility and comparison

Useful commands:

mlflow ui

Documentation:

• https://mlflow.org/docs/latest/index.html
• https://youtu.be/qdcHHrsXA48

---

MLflow with DagsHub

Tracking URI:

https://dagshub.com/vivek34561/kidney_disease_classification.mlflow

Set environment variables:

export MLFLOW_TRACKING_URI=https://dagshub.com/vivek34561/kidney_disease_classification.mlflow
export MLFLOW_TRACKING_USERNAME=vivek34561
export MLFLOW_TRACKING_PASSWORD=your_token_here

Then run:

python main.py

---

DVC Commands

dvc init
dvc repro
dvc dag

• dvc repro runs the entire ML pipeline
• dvc dag visualizes pipeline dependencies

---

🧠 Why MLflow & DVC?

MLflow

• Production-grade experiment tracking
• Parameter, metric, and artifact logging
• Model versioning and comparison

DVC

• Lightweight pipeline orchestration
• Reproducible experiments
• Data and model version control
• Ideal for PoC and research-to-production workflows

---

🔮 Future Improvements

• CI/CD integration for ML pipelines
• Automated model promotion rules
• Cloud-based artifact storage
• Model drift detection
• API-based inference service

---

👨‍💻 Author

Vivek Kumar Gupta AI Engineering Student | ML & MLOps Enthusiast

GitHub: https://github.com/vivek34561

LinkedIn: https://linkedin.com/in/vivek-gupta-0400452b6

Portfolio: https://resume-sepia-seven.vercel.app/

---

📄 License

MIT License © 2025 Vivek Kumar Gupta

• Align it directly with ML Engineer / MLOps Engineer job descriptions
• Create a system design diagram explanation for interviews