Core Concepts & Architecture

Modern AI systems have evolved from simple predictive models to complex architectures capable of multi-step reasoning. The current state-of-the-art relies on:

Transformers & Attention Mechanisms: The foundational architecture using self-attention to weigh the importance of different tokens in a sequence, enabling long-range dependency modeling.
Mixture of Experts (MoE): A sparse architecture where only a subset of parameters (experts) is activated per token, allowing for massive model sizes (trillions of parameters) with efficient inference compute.
Reasoning/Thinking Tokens: A paradigm shift where models are trained (often via Reinforcement Learning, such as GRPO) to output internal "thought" chains before providing a final answer, significantly improving performance on complex logic, math, and coding tasks.
State-Space Models (SSMs): Emerging alternatives to Transformers that offer linear scaling with sequence length, potentially enabling context windows in the millions of tokens.

Ecosystem & Product Integrations

The AI ecosystem has moved toward "Agentic Workflows," where models are integrated into IDEs (e.g., Cursor, VS Code Copilot), cloud environments (Azure AI, AWS Bedrock), and autonomous agent loops that can execute code, browse the web, and manage multi-step tool calls.

Your life does not get better by chance, it gets better by change.- Jim Rohn

AI/ML Newsletters

A live updating list of good AI/ML Newsletters | swyx

Materials

Shan Carter - OpenVisConf 2018 LESSONS FROM A YEAR OF DISTILLING MACHINE LEARNING RESEARCH

How to read machine learning research-papers? \| Applied AI Course Bianca Aguglia 1. Fast scan of the paper. 2. Careful reading. Trying to understand the logic of the paper using your current level of knowledge. 3. Working out the things not understood in step 2. 4. Question the choices made by the author of the paper. Argue with the paper. 5. Full fledged understanding. Implement the paper. Read the abstract. What problem is the paper trying to solve? Read other sources (blog posts, Wikipedia, etc) to understand the tools and techniques used in the paper. ** Mark the things you don’t understand. Decide if it’s a paper you really want to understand so you can implement it. If yes, move on to step 3. Work out the advanced math. Read other research papers. Take notes and work out the problems yourself. Explain the paper to someone else (or yourself) Are there better choices? Are there errors in the paper?

How to Read a Research Paper Ever wondered how I consume research so fast? I'm going to describe the process i use to read lots of machine learning research papers fast and efficiently. It's basically a 3-pass approach, i'll go over the details and show you the extra resources I use to learn these advanced topics. You don't have to be a PhD, anyone can read research papers. It just takes practice and patience.

Research to Code - Machine Learning tutorial A lot of times, research papers don't have an associated codebase that you can browse and run yourself. In cases like that, you'll have to code up the paper yourself. That is easier said than done, and in this video i'll show you how you should read and dissect a research paper so you can quickly implement it programmatically. The paper we'll be implementing in this video is called Neural Style transfer, that applies artistic filters to an image using 3 loss functions. Its a great starting point, i'll demo it using code, animations, and math. Enjoy!

Elementary Basic - fun book to learn programming

Reading Material & Glossary

Contents

Core Concepts & Architecture

Ecosystem & Product Integrations

AI/ML Newsletters

Materials

Elementary Basic - fun book to learn programming

Navigation menu

Personal tools

Namespaces

Variants

Views

More

Search

Navigation

Tools