MAR 18, 2026
Building a Production AI Inference Service from Scratch
How I went from a raw ONNX model to a caching, auto-batching inference gateway - and the bugs that nearly broke everything along the way.
WRITEUPS / NOTES
Blog
MAR 02, 2026
LoRA Adapters: Teaching Old Models New Tricks
How a surprisingly simple matrix trick lets you fine-tune billion-parameter models on a single GPU - without losing your mind (or your money).