Best Tutorials + Resources to Build from Scratch, Not Copying Code

Built for programmers who learn by doing, not watching or reading. Tutorials cover key skill sets in computer vision, ML + AI, and data engineering.

In this article, I share unique project tutorials designed to help you build projects alongside the tutorials, not by copying code, but by developing the actual idea yourself. Each tutorial here teaches you how to build things natively, avoiding heavy reliance on external APIs.

The Flask Mega-Tutorial: Your Path to Full-Stack Development

I know how to code and have some really cool ideas, but I often struggle to turn code into a usable platform. Enter the Holy Grail that takes you from “I don’t know how to build anything” to “Buy my B2B SaaS product that I built in the last two days”: The Flask Mega-Tutorial

Ideal for those who don’t have the patience to sit through another boring tutorial teaching you to copy code, this tutorial is structured to help you build the boilerplate for the project idea you actually want to create. In my opinion, this is the best and quickest way to transition from coding in notebooks to building full-stack applications. It is ideal for anyone with some knowledge of Python, taking you from a coder to a full-stack developer in a single walkthrough.

The best part is that it explains what each piece of code does, how it works, and why we use it in a concise but effective manner. By the end of it, you’ll be able to walk anyone through your code and how it works.

By the end you’ll know how to. . .

• set up your environment

• create, use, and migrate databases

• build APIs

• create basic front-ends

• connect your front-end and back-end.

• deploy your application.

From here you can learn more complex frameworks like Next.js or Django, but really this all you need.

Computer Vision

The best tutorials for learning unique and advanced skills you can use to build your own product instead of just creating a copy of someone else’s are from the Computer vision engineer on Youtube. Here are some of my favorite tutorials from his channel:

Customized Object Detection:

Train Yolov9 object detection custom data on Google Colab | 21:32 min

Custom Dataset Creation:

Emotion detection synthetic dataset | 35:24 min

Image Segmentation:

Image segmentation with Yolov8 custom dataset | 46:25 min

Facial Analysis:

Face recognition and face matching | 18:17 min

Image Classification:

Image classification + feature extraction | 21:59 min

Object Tracking:

Yolov8 object detection + deep sort object tracking | 34:32 min

AI / ML

Everyone wants to build AI, but few are actually creating AI projects. Here are tutorials that will help you learn how to build AI projects instead of just API wrappers:

Retrieval Augmented Generation (RAG):

Local Retrieval Augmented Generation (RAG) from Scratch (step by step tutorial) | 5:40:58 min

Large Language Models:

Create a Large Language Model from Scratch with Python | 5:43:40 min

Data Extraction + Pipelines

ML and AI models are only as good as the data they’re trained on. Learn how to extract data and create data pipelines with these top tutorials from the CodeWithYu channel on Youtube. Here are some of my favorite tutorials:

Streaming Data:

End to End Realtime Streaming with Unstructured Data | 2:30:31 min

Data Lakehouse:

Building Data Lakehouse from Scratch | 55:05 min

Algorithmic Trading:

Realtime Algorithmic Trading with Apache Flink | 1:41:34 min

Cool Libraries and Frameworks

• Babit Multimedia Framework (BMF): BMF is an open source framework built by ByteDance to simplify complex video and audio processing task. BMF enables efficient creation of applications for video transcoding, real-time filtering, live streaming, and more. It offers flexibility for custom module integration and supports GPU acceleration for high-performance processing

• YOLOv9: YOLO is a real-time object detection system widely used in computer vision applications. It's renowned for its speed and accuracy, capable of detecting multiple objects in images or video streams in a single forward pass of its neural network. Its ease of use, coupled with pre-trained models for common object classes, makes it accessible to developers with varying levels of machine learning expertise, enabling rapid integration of object detection capabilities into diverse projects.

• Video-LLaMA: This is an instruction-tuned audio-visual language model for video understanding. It makes so that LLMS are capable of understanding both visual and auditory content in a video.