Typorithm learning path
AI · LLM
Large language models — engineering with the Claude SDK and how transformers work inside.
Your learning path
0 of 24 topics completed
Saved on this device
Unit 1 · 3 topics
Prompt Engineering
System and user prompts, few-shot and chain-of-thought, and reliable structured output via tool schemas.
Unit 2 · 3 topics
Context & Tokens
Tokens and the context window, prompt caching for cost and latency, and streaming responses.
Unit 3 · 3 topics
Tools & Agents
Tool/function calling, the agent loop driven by stop_reason, and connecting tools via MCP.
Unit 4 · 3 topics
RAG & Embeddings
Embeddings and similarity, chunking into a vector store, and retrieval that grounds the model in your data.
Unit 5 · 3 topics
Tokenization & Embeddings
How input becomes vectors — byte-pair tokenization, the embedding lookup, and positional encoding with RoPE.
Unit 6 · 3 topics
Attention & Transformers
The transformer core — scaled dot-product self-attention, multi-head attention, and the residual transformer block.
Unit 7 · 3 topics
Training & Alignment
How models are made — next-token pretraining, fine-tuning with RLHF, and scaling with mixture-of-experts.
Unit 8 · 3 topics
Inference & Sampling
Generating text — temperature and top-p sampling, the KV cache and autoregressive decoding, and context cost and latency.