Build & Deploy ML Regression model with FastAPI, MLFlow, Docker, & AWS скачать в хорошем качестве

Build & Deploy ML Regression model with FastAPI, MLFlow, Docker, & AWS

This is the 2nd End-to-End project of this series. The aim is to build and deploy machine learning (and deep learning) models and focus on the whole pipeline (cleaning, training, serving layer, Docker container, CI/CD, pipeline, deploy and monitor, not just the traditional tutorials with everything done in a notebook. PS: If you're following along and deploying on AWS, I advise you to use the free tier to avoid costs or be willing to spend a few dollars on this project. If it's not within the project, then you can skip the deployment part and watch how it's done. Enjoy :) Project overview: This is a housing regression machine learning project where I analyse US housing data between 2012 to 2023. The goal is to train an ML model to predict the housing prices and deploy it to be usable by the end-user. Tech stack and tools: Great Expectation (data quality) FastAPI (HTTP endpoint) Docker (containerization) MLFlow (ML experiment tracking) GitHub Actions (CI/CD, run, test, deploy) AWS ECR, ECS (Fargate), and ALB ------- Repo & Slides https://github.com/anesriad/Regressio... https://housing-regression-mle-jpudo2... Datasets: https://www.kaggle.com/datasets/sheng... https://simplemaps.com/data/us-metros ------- TIMESTAMP: 0:00 - Intro to the project 1:25 - Pre-requisites 3:58 - Project overview / Tech stack 14:00 - Initial setup (code start) 23:50 - Notebooks (data split, cleaning) 40:28 - Notebooks (Feature engineering) 49:20 - Notebooks (ML baseline, XGBoost, finetuning) 56:34 - Pipelines (feature, training, inference) + Unit/smoke tests 01:30:32 - AWS Setup (CLI + S3) 01:38:55 - API: FastAPI Endpoints 01:46:53 - Containers: Docker 01:54:28 - UI: Streamlit 02:02:35 - MLOps (CI/CD with GitHub Actions) 02:11:00 - MLOps (Deployment on AWS ECS + ECR + S3 + ALB) ------- Next: YouTube Comment analytics for podcast recommendation (for a famous podcast), stay tuned! Add any suggestions in the comments or DM me on LinkedIn:   / riadanas