Understand the systems behind the API.

Start here

New to language models? Begin with the foundations lesson.

Already know the basics? Explore Transformer Architecture Explained: Trace a Decoder-Only LLM

Start Here: LLM Foundations · 8 min readYour First LLM Walkthrough: From Prompt to AnswerFollow a small prompt through token choices and explain the repeated step that produces an answer.Read the lesson

Inference primers

Two short pieces that set up the vocabulary before the numbered course.

• How LLM inference works9 min read
• Understanding the KV Cache12 min read

Learning tracks

4 tracks have lesson drafts available; the other tracks remain planned.

• Track 00In progress

  Start Here: LLM Foundations

  Six guided parts take a new reader from a raw prompt through the math, transformer, generation phases, cache, and GPU basics. Every new term is explained before it is used in depth.Part I — Foundations: What Happens When You Send a Prompt? · Part II — Mathematical Foundations for Transformer Inference · Part III — Inside a Decoder-Only Transformer · Part IV — Prefill and Decode: The Two Major Phases · Part V — KV Cache: From First Principles to Advanced Techniques · Part VI — GPU Fundamentals for Inference20 of 24 chapter drafts available
• Track 01In progress

  LLM Internals & Architecture

  Implement and inspect a decoder-only transformer, explain how it learns and generates, and evaluate the design choices behind modern and reasoning models.Transformer architecture · Tokenisation and embeddings · Attention and masking · Positional encoding · Pretraining and generation · KV cache and context · MQA, GQA, and MLA · Mixture of Experts · Multi-token prediction · Architecture comparison · Reasoning evaluation · Reasoning distillation2 of 2 primers · 26 of 26 chapter drafts available
• Track 02In progress

  Inference Optimisation & AI Infrastructure

  Make models fast and affordable to serve: memory layout, batching, quantisation, and the GPU behaviour behind every latency number.Part VII — Attention optimization and efficient kernels · Part VIII — Batching and inference scheduling · Part IX — Model compression and faster decoding · Part X — Multi-GPU and distributed inference · Part XI — Production inference engines · Part XII — Performance engineering and capacity planning · Part XIII — Hands-on implementation · Part XIV — Advanced systems design and engineering trade-offs16 of 35 chapter drafts available
• Track 03In progress

  Fine-Tuning & Domain-Specific Models

  Adapt a base model to a domain without losing the judgement you started with: datasets, adapters, distillation, and small models.Dataset engineering · Supervised fine-tuning · LoRA and QLoRA · Distillation · Small language models16 of 16 chapter drafts available
• Track 04Planned

  RAG & Agent Systems

  Build systems that retrieve, decide, and act, with retrieval quality, orchestration, and failure handling treated as first-class engineering problems.Retrieval · Reranking · Orchestration · Tool execution · Memory · Multi-agent coordinationChapters coming soon
• Track 05Planned

  AI Evaluation, Reliability & MLOps

  Turn “it seems better” into evidence: evaluation sets, observability, versioning, and release gates that survive real traffic.Evaluation frameworks · Observability · Versioning · Deployment · Regression gates · GovernanceChapters coming soon
• Track 06Planned

  Reinforcement Learning & Self-Improving AI

  Understand how models learn from feedback: reward design, preference optimisation, and the loops that make agents improve without drifting.Reward design · Preference optimisation · RLHF · Feedback loops · Agent improvementChapters coming soon
• Track 07Planned

  System Design & Distributed Systems

  The durable systems skills behind every AI product: storage, queues, consistency, concurrency, and capacity under failure.Databases · Storage · Caching · Messaging · Consistency · Concurrency · Capacity planning · Reliability · MigrationsChapters coming soon
• Track 08Planned

  Staff+ Engineering: Judgment, Strategy & Leadership

  The work that compounds beyond code: framing decisions, writing clearly, aligning teams, and connecting engineering to the business.Architectural decisions · Technical writing · Cross-team execution · Technical strategy · Influence · Mentoring · Business alignmentChapters coming soon