AI Foundations for Engineers
How modern AI works, from neural networks and transformers to RAG, fine-tuning and agents.
- Lesson 1
Start here: AI Foundations for Engineers
A map of how a modern AI system fits together, from training a transformer to serving it in production, and where each lesson in the series goes deeper.
- Lesson 2
Neural Nets & Embeddings
How neural networks represent meaning as embeddings, how backpropagation adjusts millions of weights, and why training needs far more memory than inference.
- Lesson 3
Transformers & Attention
How attention lets a transformer relate every token to every other, why that cost grows quadratically, and the tricks, like the KV cache, that keep it fast.
- Lesson 4
Scaling, RLHF & Reasoning
What scaling laws say about model size and data, how RLHF shapes a model's behaviour, and where today's reasoning models come from, risks included.
- Lesson 5
RAG & Vector Search
Building advanced language model applications requires overcoming the limitations of naive retrieval through sophisticated preprocessing and search architectures.
- Lesson 6
Quantization & LoRA
Why large language models hit a memory wall, and how quantisation and LoRA shrink models and fine-tuning so they fit and run fast on real GPUs.
- Lesson 7
Agents, Tools & MCP
How AI agents plan and call tools, how the Model Context Protocol (MCP) connects them to data and services, and when a plain workflow beats an agent.