Artificial intelligence has acquired a vocabulary of its own: machine learning, neural networks, tokens, embeddings, transformers, pre-training, RLHF, RAG, fine-tuning. The words often arrive faster than the explanations.
This series is meant to slow that down.
You do not need a machine-learning background to begin. We will start with the relationship between AI, machine learning, and deep learning, then follow the path all the way to the systems behind modern language-model applications.
The goal is not to make every topic sound simple. It is to make each topic understandable without making it inaccurate.
What you will understand by the end
By the final lesson, you should be able to explain:
- how AI, machine learning, and deep learning relate;
- how text, images, and audio become numerical representations;
- the differences among supervised, unsupervised, self-supervised, and reinforcement learning;
- what neural-network weights actually do;
- what a training corpus contributes and what it does not;
- how transformers use attention and predict the next token;
- how forward passes, loss, backpropagation, and optimization fit together;
- why large training runs consume so much compute;
- how post-training turns a base model into a useful assistant;
- what RLHF and RLAIF can improve, and what they cannot guarantee;
- why RAG can ground answers without eliminating hallucinations;
- when prompting, retrieval, and fine-tuning are the right tools.
How to use the series
Read it in order if this is your first serious introduction to AI. The later lessons assume the mental models built in the earlier ones.
If you already work with AI systems, use the module list as a reference. Each lesson stands on its own and is intentionally short enough to revisit.
Module 1: The Foundations
Module 2: How Machines Learn
Module 3: Deep Learning and Large Language Models
Module 4: Inside the Training Loop
Module 5: Post-Training and Alignment
Module 6: Applying Large Language Models
After the fundamentals
This series explains the foundations. The rest of jacobpallattu.com explores what happens when those models become part of real software systems: retrieval, tools, agents, governance, architecture, evaluation, and experimental model work.
When you finish the final lesson, you can continue into PallattuLM, the site's experimental language-model work, or browse AI-native systems and architecture.