Build local AI with Ollama, Open WebUI and RAG. In three days.
Get your own pipeline running and explore prompt injection in a hands-on hackathon.
- 3 days
Hands-on workshop
- Ollama and Open WebUI
Local models and a web interface
- Python and RAG
Build a pipeline
Who this workshop is for
For teams that want to work with local language models and see how a small AI application fits together. You’ll run a model, set up an interface and write code. We’ll agree on prerequisites and the technical setup before you book.
If your team needs a shared foundation, start with the generative AI fundamentals course. For a deeper look at APIs, integration and deployment, explore AI training for software developers.
From running a model to building your own pipeline
We adapt the order and depth of the exercises to your group.
Use Ollama
Run a local language model and see how it works in the workshop environment.
Set up Open WebUI
Install the interface and use it to interact with the model.
Write Python
Use Python in the hands-on exercises and in the pipeline you’ll build.
Build a RAG pipeline
Connect documents, retrieval and model output in one small exercise.
Understand prompt injection
Explore how inputs can affect instructions and where system prompts have limits.
Test what you’ve built in a short hackathon
Use the practice system to explore where its instructions hold up and discuss what you observe.
What you’ll do in the workshop
Use a model with Ollama.
Set up Open WebUI and interact with the model.
Use Python to build a RAG pipeline.
Run and discuss prompt-injection tests in a practice environment.
What you can take on independently afterward depends on your experience and practice. Production use requires its own assessment and approval.
Look critically at how the system behaves
In the hackathon, you examine how a practice system handles instructions and access to protected information. We agree on the environment and permitted materials in advance. Together we review what the tests show and which questions remain open.
Data privacy also depends on how you configure a local system. Before training, we agree on models, features and where data flows. Our guide to data sovereignty for AI covers those decisions in more detail.
What we agree on up front
Exercises and setup tailored to your group.
Roles and experience
Who’s taking part, their Python and AI experience, and how far you want to go.
Technical setup
Devices, installation, access and approved models. We agree on what to prepare before the workshop.
Exercises and data
Examples, available documents and rules for the practice environment.
Format and scope
Three workshop days. Your proposal covers dates, group size, materials, support and price.
He builds what he teaches.
:quality(80))
Martin Kogut
Founder, AI consultant and trainer
Martin Kogut is a software developer and the founder of hypescale in Berlin. He teaches teams how to work with local language models, Ollama, Open WebUI and Python, build RAG pipelines and assess prompt injection.
- Local LLMs
- Ollama
- Open WebUI
- Python
- RAG
- Prompt injection
Questions about the local LLM workshop
What to explore next
To work with company knowledge, explore AI knowledge bases and RAG. For a broader learning programme, see all AI training courses. For an example of a team moving from training to evaluating real tools, see the Carano case study.
Let’s plan a workshop for your team.
Tell us who’s taking part, what they already know and what tools they can use. We’ll tailor the three days to your group.