Talki Academy
🧠

Understanding LLMs: A Complete Foundation

An L100 foundations course. Before calling an API or launching an agent, understand what is under the hood. This conceptual, visual course demystifies large language models: how a Transformer works, how it is trained, what hardware it runs on, and how much it costs. No code required — you leave with a solid understanding to make informed AI decisions.

Duration
2 days
Level
Beginner
Price
9.99 EUR/month (all courses included)
Max group
15 participants

What you will learn

+Explain how a neural network and deep learning work
+Describe the Transformer architecture and the attention mechanism
+Understand the training pipeline: pre-training, RLHF, DPO
+Tell model families apart and know when to use which
+Assess hardware constraints: GPU, VRAM, quantization
+Analyze LLM economics: training, inference, and per-token costs

Course program

Module 1: Neural networks and deep learning

3h30
  • From the biological neuron to the artificial neuron
  • What a deep neural network is
  • Supervised, unsupervised, and reinforcement learning
  • Interactive demo: watch a network learn

Module 2: The Transformer architecture

3h30
  • Tokenization and embeddings: how text becomes numbers
  • The attention mechanism explained visually
  • Encoder vs decoder: BERT, GPT, and beyond
  • Scaling: why larger models work better

Module 3: Training and alignment

3h30
  • Pre-training: massive data, objectives, and costs
  • RLHF and DPO: making a model helpful and safe
  • Model families: GPT, Claude, Llama, Mistral — positioning
  • Open-source vs proprietary: selection criteria

Module 4: Hardware, economics, and ethics

3h30
  • GPU vs CPU, VRAM, INT4/INT8 quantization
  • Demo: run a model locally with Ollama
  • LLM economics: training, inference, and per-token pricing
  • Ethics: hallucinations, bias, environmental impact

Ready to get started?

9.99 EUR/month — All courses included, cancel anytime

Subscribe — 9.99 €/monthView all courses