MLOps and Deployment

Training is half the job. Deployment is how users actually get predictions. MLOps covers saving models, serving them, and keeping them healthy in production.

You will save a trained pipeline with joblib, wrap it in a FastAPI endpoint, and test it locally. Streamlit lets you build a demo UI in pure Python. Docker makes it portable.

The biggest production pitfall is not the model but the data: distributions drift, and accuracy drops. Monitoring beats retraining blindly.

Save and serve

Save the whole pipeline, not just the model, so preprocessing is included. Serve with FastAPI.

import joblib
joblib.dump(pipe, "model.joblib")
# FastAPI: from fastapi import FastAPI; app = FastAPI()
# @app.post("/predict")
# def predict(x: dict): return {"pred": pipe.predict([x["text"]])[0]}

Docker and Streamlit

Dockerfile copies code and installs deps. Streamlit turns a script into a web app.

# Dockerfile
# FROM python:3.10
# COPY requirements.txt .
# RUN pip install -r requirements.txt
# CMD ["uvicorn", "app:app", "--host", "0.0.0.0"]

Syntax

import joblib
joblib.dump(pipe, "model.joblib")
pipe = joblib.load("model.joblib")
pipe.predict([["new data"]])

# Never: joblib.load("untrusted.pkl")
# Only load your own model.joblib