Serving Complexity

Large models create complex systems problems. This blog is about solving them. Notes on distributed systems, C++, and LLM infrastructure.


How to Build a Multi-Dimensional Array by Separating View Logic from Compute Kernels

A deep dive into my multi-dimensional design. I explain the "Control Plane vs. Data Plane" model, which implements "smart" Python to handle all the complex view and stride logic, letting us build a "dumb," simple, and incredibly fast C++ and CUDA backend

7 min read   ·   2025

An Engineer's Guide to Deep Learning Optimizers

A deep-dive into deep learning optimizers where we trace the evolution from SGD to Momentum and finally to Adam.

15 min read   ·   2025

a distill-style blog post

an example of a distill-style blog post and main elements

23 min read   ·   2021

a post with code

an example of a blog post with some code

4 min read   ·   2015