🩺 Diabetes Prediction Model – MLOps Project (FastAPI + Docker + K8s)

This beginner-friendly project walks you through building and deploying a machine learning model for diabetes prediction. You’ll learn how to:

• ✅ Train a model using Random Forest
• ✅ Serve it with FastAPI
• ✅ Dockerize the application
• ✅ Deploy it on Kubernetes (with Kind)
• ✅ Understand CI/CD and monitoring steps for future improvements

---

📊 Problem Statement

Predict if a person is diabetic based on:

• Pregnancies
• Glucose
• Blood Pressure
• BMI
• Age

Model used: Random Forest Classifier
Dataset: Pima Indians Diabetes Dataset

---

🚀 Quick Start

1. Clone the Repository

git clone https://github.com/iam-veeramalla/first-mlops-project.git
cd first-mlops-project

---

💻 Setup and Model Training

2. Create a Virtual Environment

🔹 Windows (CMD)

python -m venv .mlops
.mlops\Scripts\activate

🔹 Windows (PowerShell)

python -m venv .mlops
.\.mlops\Scripts\Activate.ps1

🔹 macOS/Linux

python3 -m venv .mlops
source .mlops/bin/activate

🔹 Windows with Git Bash

python -m venv .mlops
source .mlops/Scripts/activate

3. Install Python Dependencies

pip install -r requirements.txt

4. Train the Model

python train.py

➡️ This will generate a model.pkl file used for predictions.

---

🌐 Run FastAPI Locally

5. Start FastAPI Server

uvicorn main:app --reload

Open your browser: http://localhost:8000/docs

6. Sample Input for /predict endpoint

{
  "Pregnancies": 2,
  "Glucose": 130,
  "BloodPressure": 70,
  "BMI": 28.5,
  "Age": 45
}

---

🐳 Dockerize the API

7. Build Docker Image

docker build -t diabetes-prediction-model .

8. Run the Docker Container

docker run -p 8000:8000 diabetes-prediction-model

9. Tag Docker Image

docker tag <image-id> iamvikramkumar/diabetes-prediction-model:v1

(Find <image-id> using docker images)

10. Push to Docker Hub

docker login
docker push iamvikramkumar/diabetes-prediction-model:v1

---

☸️ Deploy on Kubernetes (Using Kind)

📝 Note: Please install Docker Desktop, Kubectl, and Kind before proceeding.

11. Create a Kind Cluster

kind create cluster --name demo-mlops

12. Check Cluster Status

kubectl config current-context
kubectl get nodes

13. Apply Kubernetes Manifest

kubectl apply -f deploy.yaml

14. Monitor Kubernetes Resources

kubectl get pods -w
kubectl get svc

15. Port Forward to Access API

kubectl port-forward svc/diabetes-api-service 1111:80 --address=0.0.0.0

🔗 Open: http://localhost:1111/docs

---

🔄 Future Scope: CI/CD and Monitoring

🧪 CI/CD (GitHub Actions, Not Implemented Yet)

• Auto test and validate model updates
• Build Docker images
• Push to Docker Hub
• Deploy to Kubernetes

📈 Monitoring (Not Implemented Yet)

• Prometheus + Grafana for API performance
• Evidently AI or WhyLogs for data drift and prediction accuracy
• Alerts for latency or failure spikes

---

🙌 Credits

Created by ABHISHEK VEERAMALLA

Subscribe for more DevOps + MLOps content on the YouTube Channel - Abhishek.Veeramalla