How language models work
Tokens, context, hallucinations and model choice: the basics to use LLMs knowingly and write prompts that work.
- Audience:Business and technical
- Duration:8 hours · 1 day
- Format:On-site or online
- Language:Italian or English
What you'll learn
- Explain in plain words how an LLM produces an answer
- Understand why a model gets things wrong and how to reduce hallucinations
- Choose between commercial and open-source models on cost, quality and privacy
- Write structured, reusable prompts
- Evaluate model answers systematically
Course content
4 modules · 8 hours in total
1. Inside an LLM2 hours
- Tokens, context window, temperature
- Training, fine-tuning and inference
- Why a model does not "know" what it was not given
2. Choosing the model2 hours
- Commercial and open-source models compared
- Costs, latency and where the data ends up
- Small models for specific tasks
3. Prompt engineering2 hours
- Role, context, examples and output format
- Reusable prompts for the team
- Exercises on your documents
4. Evaluating answers2 hours
- Hallucinations: how to spot and limit them
- Test sets and quality criteria
- When an LLM is not the right tool
Requirements
- Having used a chatbot such as ChatGPT, Claude or Gemini at least once
Description
Using a chatbot is easy. Understanding why it sometimes answers brilliantly and sometimes makes things up is another matter, and it is the difference between occasional use and use you can rely on.
This day explains how language models work without formulas, through examples and exercises. It is the shared foundation for people who want to use them at work and for teams who will design AI systems.
Who this course is for
- Professionals in any area who use or will use LLMs
- Product, marketing and customer care teams
- Anyone planning to take the advanced courses on RAG, agents and security
Request information
The course is not sold online: we run it for your team. Write to us and we will send you the programme and a proposal.
More courses in this category
TechnicalLanguage models
RAG and AI agents: assistants built on company data
Design, evaluate and ship assistants that answer from your documents and agents that use your systems.
- 16 hours · 2 days
- On-site or online
See the programme
TechnicalLanguage models
Securing LLM systems: AI red teaming and guardrails
How AI systems get attacked and how to defend them, with live red teaming and what the AI Act, Cyber Resilience Act and NIS2 require.
- 16 hours · 2 days
- On-site or online
See the programme