This hub tracks the system layer around modern AI.
Linked Notes
- About for my broader systems interests
- Build GPT-2 from Scratch for implementation details
- Deep Learning Explained with Mathematics for the mathematical view behind training and inference
- A Simple CPU-based Whirl Shader Experiment for low-level computational experimentation
- Transformers for architecture-level design choices
- Mathematical Foundations for probability, optimization, and learning dynamics
Questions
- how architecture and inference constraints shape model design
- how systems thinking changes the way we study AI
- where correctness, latency, and scale collide
Bridges
- from Transformers into memory layout, KV cache, and serving constraints
- from Mathematical Foundations into optimization stability and uncertainty-aware systems
- from Computational Graphics into performance-aware rendering and visual computation