ML Model Deployment & MLOps with FastAPI, Streamlit, MLflow

Welcome to ML Model Deployment & MLOps with FastAPI, Streamlit, MLflow course. This is a comprehensive project based course where you will learn how to build machine learning models, deploy the model, monitor the model performance and also retrain the model using new data. This course is a perfect combination between python and machine learning, making it an ideal opportunity to practice your programming skills while improving your technical knowledge in data science.

In the introduction session, you will learn the basic fundamentals of machine learning model deployment and machine learning operations, such as getting to know common tools that are frequently used for ML model deployment and MLOps workflow. Then, in the next section, we will download earthquake and flight price datasets from Kaggle, it is a platform that provides many high quality datasets from various industries. Afterward, we are going to build two machine learning models.

Firstly, we are going to build an earthquake detection model using Random Forest Classifier. This model will be able to predict earthquake alert level based on features like magnitude, community intensity, and significance level. Following that, we are also going to build a flight ticket price prediction model using Decision Tree Regressor.

This model will be able to predict ticket price based on features like airline, destination cities, number of stops, departure time, arrival time, and flight duration. Then, in the next section, we are going to deploy the machine learning model using various frameworks. We are going to create a full interactive web interface using Gradio, Streamlit, Flask, and Dash.

After that, we are going to deploy the machine learning model on Hugging Face space where you can host and share your model. In addition, we are also going to deploy the machine learning model using FastAPI, which will enable users to send prediction requests through an API endpoint. Then, in the next section, we are going to track model performance by displaying metrics like accuracy, precision, recall, and F1 score using MLflow.

Additionally, we are also going to package our trained model into a reproducible format, this will make it easier to deploy across different environments. Then, after that, we are going to perform data augmentation, specifically, we are going to generate synthetic data using Scikit Learn and this new data will be used for retraining the machine learning model. In addition to that, we are going to monitor data quality by checking missing values, duplicates, and outliers in the synthetic data.

Lastly, at the end of the course, we are going to retrain the machine learning model using Apache Airflow. Firstly, before getting into the course, we need to ask this question to ourselves, why should we learn about machine learning model deployment and operation? Well, here is my answer, no matter how good your machine learning model is, there is no use if nobody can access or try it.

That is the reason why understanding how to deploy your model is very important, so you can share your model and enable users or other developers to interact with your model in real time. Once the model is deployed, your job is not done, as the model is retrained with new data, there might be potential for data drift, drop in accuracy, and performance decline which is why monitoring and maintenance are very essential.