Sebastian Raschka
Sebastian Raschka is a machine learning researcher and educator known for his accessible technical writing and his ability to explain frontier AI concepts clearly without sacrificing depth.
Notable Work
- Build a Large Language Model From Scratch (manning-publications, 2024) — step-by-step implementation of a GPT-like LLM in PyTorch from first principles: tokenization, attention-mechanism, gpt-architecture, and pretraining. See raschka-2024-build-llm-from-scratch.
- Independent educator and writer; publishes technical breakdowns of AI papers and techniques
- ML researcher with hands-on experience in model training
Views on Architecture
Raschka’s consistent position: the transformer architecture is fundamentally unchanged from GPT-2. Modern frontier models are autoregressive decoder-only transformers with attention + FFN blocks — the core is the same; the innovations are engineering tweaks for efficiency and capacity (mixture-of-experts, GQA, MLA, RMSNorm, SwiGLU).
He views the “AGI” concept as continuous improvement (“amplification not paradigm change”), not a step-function transition.
Sources: fridman-lambert-raschka-2026-state-of-ai | raschka-2024-build-llm-from-scratch