🧠

LLM Foundations

Transformer layers, tokenization, training stages, MoE vs Dense models.

1

What is an LLM & How It works

beginner~8h

Learn the core neural architectures, conditional probability, tokens, and dense vs MoE configurations.

2

Training Stages: Pretraining, SFT, RLHF, GRPO

intermediate~12h

Deconstruct the four stages of building a production reasoning LLM from scratch.

3

Text Generation: Decoding & Sampling

beginner~6h

Learn generation parameters (temperature, top-p, top-k) and local execution runtimes.

4

7 LLM Generation Parameters

beginner~3h

The 7 levers that control every LLM generation — max tokens, temperature, top-k, top-p, frequency penalty, presence penalty, and stop sequences — plus bonus min-p sampling.

5

4 LLM Text Generation Strategies + SLED

intermediate~3h

Greedy, Multinomial Sampling, Beam Search, and Contrastive Search — the 4 core strategies for picking the next token — plus SLED (Self-Logits Evolution Decoding), a bonus technique that improves factuality using every layer's predictions.

6

3 Techniques to Train an LLM Using Another LLM (Distillation)

advanced~4h

Soft-label, hard-label, and co-distillation — the three ways real production LLMs (Llama 4, Gemma, DeepSeek) train a smaller student model using a larger teacher.

7

4 Ways to Run LLMs Locally

beginner~2h

Ollama, LM Studio, vLLM, and llama.cpp — the four most common ways to run LLMs on your own machine, compared for privacy, ease of use, and performance.

8

Mixture of Experts: Router Training Challenges & Solutions

advanced~3h

Why naively training a Mixture-of-Experts router causes expert under-training, and the two concrete fixes (noise injection + top-K masking, token capacity limits) that solve it.