AI educator and LLM research engineer Sebastian Raschka has released a full video course on building reasoning models from scratch. The course spans eight coding videos and more than 11 hours of content, all built in Python and PyTorch, according to his Ahead of AI magazine.
The course starts from a pre-trained LLM and walks through methods to evaluate and improve it, covering evaluation, inference-time scaling, reinforcement learning and distillation. Raschka describes it as the sequel to his earlier Coding LLMs from the Ground Up course, though prior experience with that course is not required.
What the reasoning models course covers
Raschka has written extensively about reasoning models over the past two years. In his announcement, he notes that almost all current-generation LLMs are either reasoning models or offer a reasoning mode, making the topic more relevant than ever.
The new video series accompanies his written Build Reasoning Model (From Scratch) guide, which implements the core techniques from the ground up in PyTorch. Raschka says this hands-on approach remains the best way to understand how reasoning models really work.
- 8 coding videos, over 11 hours of content
- Evaluation of pre-trained LLMs
- Inference-time scaling techniques
- Reinforcement learning for reasoning
- Distillation methods
