Learning Together to Build LLMs at Scale with Open-Source Coursework

In this presentation, we will share our experience of bringing together a study group to learn how to build and scale language models from scratch using Stanford’s open-source CS-336 coursework.

What started as a straightforward plan with weekly study sessions and hands-on homework, quickly came face to face with reality: varied skill levels, competing priorities, and schedules that didn’t resemble a university semester. Along the way, we discovered that organizing and sustaining the group was at least as challenging as understanding the trade-offs in scaling and optimizing the training of large language models.

In this session, we’ll discuss the motivations and tactics that helped us form and keep an engaged learning community. We’ll cover how we adapted the material for participants with diverse backgrounds, restructured expectations when the original plan fell short, and leaned into individual members’ expertise to create a more sustainable and rewarding experience.

The talk is practical, honest, and intentionally humorous. Attendees will leave with actionable insights for running their own technical study groups, learning effectively from open-source coursework, and navigating the inevitable “bumps in the road” that come with expanding their knowledge, especially when doing so alongside others.

• AI/ML
• Soft Skills