Build Your Own LLM - Full-Day Workshop
Want to learn how to build your own large language model from scratch? Join us for a free full-day workshop in San Francisco.
What You'll Learn
Our goal is for you to develop an intuition of LLMs by deep-diving into every aspect: coding, frameworks, demos, and math. The workshop will cover 22 topics across four milestones, each with its own deliverable:
Intro: Get a working off-the-shelf GPT-2 style transformer up & running.
Deep Neural Networks: Train a basic neural net from scratch and learn machine learning fundamentals as we do that.
Transformer Architecture: Artisanally hand-code every component of a transformer: tokenizer, embeddings, attention, MLP, normalization, regularization, etc. By the end of this module you'll load pretrained weights into your model and have it generate text.
Train Your Own LLM: Train, evaluate, instruct fine-tune and RL a language model. Walk away with your own model named after you.
Every topic alternates ~10 minutes of teaching with ~10 minutes of hands-on exercises. See one, do one. We'll have interactive demos, coding samples, and math-by-hand as exercises.
Who This Is For
No prior knowledge of calculus, linear algebra, python/pyTorch, or machine learning is required. It's totally fine and even encouraged to not know any of that stuff prior to showing up.
We do ask three things:
You're extremely comfortable with learning new concepts through code, demos, and exercises. This workshop is aimed at people who code.
You can commit and stay focused for an 8-hour knowledge-packed Saturday workshop. It's going to be a lot.
You bring a laptop with an active Google Colab subscription ($9 paid, free with .edu email).
Schedule
Saturday, May 2nd, 2026 9:00 AM – 5:00 PM The Commons: 540 Laguna St, San Francisco
We'll move through 22 topics at ~20 minutes each. There's a 30-minute lunch break around noonish. When you sign up, you have the option of joining a group order for Banh Mi from Dragon Eats (vegan/vegetarian options available) or bring/buy your own lunch.
Your Instructor
Justin Angel has been a software engineer for 23 years. His last role was as an L8 SWE at Meta. He's worked at Meta, Uber, Amazon, Apple, and Microsoft building apps, developer platforms, and hardware. He recently completed his Master's in AI/ML, writing his thesis on LLM psychotherapy. He picked up AI/ML in the last six months, and designed this workshop to share his knowledge of LLMs with you.
Other Details
We're going to order lunch from Dragon Eats (Vietnamese). We'll collect payment via Venmo day of the event.
Limited to 30 spots so everyone gets individualized support as needed. Approval required for participation.
Questions?
Reach out to Justin directly at [email protected] or Twitter DM @JustinAngel. You can review the training materials (WIP) @ go.JustinAngel.ai/deck
Presented by Rally SF, a community for AI builders in San Francisco. More events on Rally's Luma Calendar.