2.5 KiB
Bio
Reese is DIY technology enthusiast with a passion for projects that make things easy. He's been working in development since 2017 with experience in risk, compliance, scripting automation, full stack web development, container infrastructure, homelab server hardware, ESP Home and home automation. Reese has a passion for mentoring, but even more of a passion for sharing the new tech he found last week with anyone who will listen. Reese wants tech to be fun and approachable for anyone at any skill level.
Credentials
Reese has spoken at multiple company conferences about building websites and automating with Python. He's taught multi-day intro-to-python classes both online and in person. He has 8 years of industry experience, 3 of which have been spent growing the development team at a Nimbis Services. Reese has, professionally and personally, written and distributed Python pip packages, designed and hosted websites, built and deployed a version control system, led AI development teams, taught Python classes, mentored high school students in tech, annoyed his friends with discord bots, and automated his bathroom fans. He's accustomed to speaking in front of large and small audiences and relishes the opportunity to share his excitement with a crowd.
Abstract
This talk will walk through the process of putting your local LLM to good* use. Through the medium of a Discord bot, we will explore how to leverage llama.cpp to give your friends the ability to create custom bots with custom personalities, have those personalities talk with each other, generate images, edit images, and set yourself up to leverage tool calling so your bots can interact with the real world.
We will cover the state of hosting offline LLMs and discuss some strategies for hosting them safely with Podman, Bifrost, and Caddy. We will also discuss the current state of LLM hardware and give some realistic examples with AMD, Intel, Nvidia, and CPU based solutions. We will not be using cloud examples, as this talk will focus on avoiding cloud solutions in general. We will poke fun at leveraging Discord as our example if our goal is to self-host.
Ultimately, I want this talk's participants to leave with some functional code and good ideas to get them thinking about ways they can integrate LLMs into their communities while maintaining control and privacy (and avoiding a hefty bill). This talk will emphasize audience participation to generate ideas for a prebuilt demo of the custom bot service, but will not build anything live during the presentation.