TC
English

Technical writing

Technical articles on AI engineering

Code-first technical articles where the implementation, assumptions, metrics, and limitations stay visible. Conceptual tutorials will appear here as they are published.

Latest

Article

What Actually Speeds Up Transformer Inference?

Profiling and optimizing a small autoregressive transformer with JAX, KV caching, batching, graph compilation, and low-bit inference.

16.23x Best same-batch speedup 6,225 tok/s Best throughput
JAX Inference KV cache Profiling
Read article

More technical writing

Article

Training a 10M-Parameter Transformer to Learn 3-Digit Arithmetic

A code-first experiment that builds a character-level decoder-only transformer with JAX, Flax, and Optax, then trains it on generated addition, subtraction, multiplication, integer division, and modulo problems.

10.65M Parameters 4,998,000 Examples
Read article