Fabian Blaicher

Launching a “train your own LLM” course next week with @fb_ldn! So excited about this one ☺️ One-liner: deeply understand how LLMs work by training one from scratch based on Karpathy’s nanochat, in a small group of technical founders/engineers, guided by an AI/ML expert — all in 3 weeks! Basically we realized that Feynman was right and you can’t fully understand something until you create (or recreate) it; hence this course. We will cover: * what’s an LLM * what are tokens, embeddings, activation functions * what’s a transformer, what is attention, how it works * what data an LLM is trained on, where this data comes from * what training means (finding weights for neurons to activate) * what is loss, why it improves with training * what is fine-tuning * what evals mean, how to evaluate an LLM after training By the end of 3 weeks, you’ll have trained a gpt2-lvl LLM from scratch and acquired an intuitive + first-principles understanding of how these models are built & trained, and what their limitations are Logistics: 3 live lectures (1hr each) + 3 office hours + async homework, small group of ~10 technical founders, myself and @fb_ldn as instructors (Fabian has an AI & ML background, co-founded Shipamax W17 and exited to WiseTech Global) DM if interested!