Sentiment Analysis – MLOps Project

This project demonstrates an end-to-end Machine Learning + MLOps workflow for sentiment analysis, covering model training, experiment tracking, and model serving using FastAPI.

The goal is to build a production-style ML service that can be trained offline and served online via a REST API.

---

🚀 Features

• Train a sentiment analysis model
• Track experiments using MLflow
• Save trained models locally
• Serve predictions using FastAPI
• REST API for real-time inference
• Clean, scalable project structure (src/ based)

---

🗂️ Project Structure

SENTIMENT-MLOPS/ ├── src/ │ ├── train.py # Model training logic │ ├── predict.py # Model loading and inference │ ├── api.py # FastAPI application │ └── init.py ├── data/ # Local datasets (ignored in git) ├── models/ # Trained models (ignored in git) ├── mlruns/ # MLflow experiments (ignored in git) ├── requirements.txt # Python dependencies ├── .gitignore └── README.md

---

⚙️ Setup Instructions

1️⃣ Clone the repository

git clone https://github.com/OLAS05/sentiment-mlops.git
cd sentiment-mlops

2️⃣ Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate   # Linux / Mac
.venv\Scripts\activate      # Windows

3️⃣ Install dependencies
pip install -r requirements.txt

🧠 Model Training
Run the training script:
python src/train.py

This will:
Train the sentiment model
Log experiments to MLflow
Save the trained model locally

To view MLflow UI:
mlflow ui
Open: http://localhost:5000

🌐 Run the API Server
Start the FastAPI app:
uvicorn src.api:app --host 0.0.0.0 --port 8000

Health Check
GET /

Prediction Endpoint
POST /predict

Sample Request
{
  "texts": ["I love this product!", "This is terrible"]
}

Sample Response
[
  {"label": "POSITIVE", "score": 0.99},
  {"label": "NEGATIVE", "score": 0.98}
]

🛠️ Tech Stack
Python
FastAPI
Hugging Face Transformers
PyTorch
MLflow
Uvicorn