Keeping Things Local: Build Private LLMs for Your Team

What good is a private AI if you can only use it from the system it runs on?

This four-hour workshop takes security practitioners past simply installing Ollama and into building something bigger: a locally hosted model your whole team can reach.

You’ll customize models for your own workflows, connect systems over an encrypted mesh network, and put an authentication layer in front of everything. Because it all runs on hardware you control, sensitive data never leaves your environment while access expands beyond a single desk.

Show up with a system that can run two virtual machines, and walk out with a working private service you built yourself.

Once you’ve built a local LLM, learn to prompt it for reliable, consistent results in the follow-up workshop Prompt Engineering with Local LLMs.

Who Should Take This Workshop

Security practitioners, sysadmins, and technical folks who want AI capability without shipping sensitive data to a third party. If you handle client data, work under confidentiality constraints, or just want to know exactly where your prompts go, this workshop is for you. It is a strong fit for:

• Red teamers and penetration testers, who get a private AI that won’t leak client data during an engagement plus mesh networking they can use on assessments
• Consultants and MSSP staff who handle multiple clients’ sensitive data under contract
• Anyone working in a regulated industry (healthcare, legal, finance, government) where data sovereignty is a compliance requirement, not a preference

What You’ll Learn

By the end of the workshop, you will be able to:

• Install and configure Ollama on Linux and manage local models from the command line
• Configure Ollama to listen on the network so other systems can use it
• Customize model behavior with Modelfiles to build purpose specific assistants for your workflows
• Build an encrypted mesh network with Tailscale and connect multiple machines to it
• Put an nginx reverse proxy with basic authentication in front of Ollama to control who can reach it
• Explain the data sovereignty case for local AI and the attack surface you take on when you expose an LLM service