Kernel Nomad

A cache hit is only part of the story

Mon, 12 Oct 2026 12:20:00 +0300

Start with a visible request path

The service lab has a short path to follow: a Python API, PostgreSQL for records, and Redis for a 60-second read cache.

Creating an item writes it to PostgreSQL. The first read retrieves it from the database and stores a cached copy. A second read can return the cached record. The response includes a source field so the route is easy to see while experimenting.

Write the question before the benchmark

Mon, 12 Oct 2026 12:10:00 +0300

What would the number mean?

Before timing a kernel, I want to write down the question that the timing is supposed to answer.

“Is this fast?” leaves a lot open. A narrower question is more useful: how does one change in the implementation affect the same workload, on the same device, with the same definition of elapsed time?

A small lab, a longer attention span

Mon, 12 Oct 2026 12:00:00 +0300

A useful division of work

The Mac is the editing surface. The Ubuntu laptop is where the code runs.

That separation makes the environment concrete. Files live on the Linux machine, builds happen there, and a CUDA program runs against the GPU that is physically in that laptop.

A small lab, by design

Mon, 01 Jan 0001 00:00:00 +0000

The setup is simple: a Mac for editing, an Ubuntu laptop for execution, and an SSH connection between them. The Linux machine is also the home of this notebook.

The machine

Part	Working setup
Operating system	Ubuntu 24.04 LTS
GPU	NVIDIA GTX 1050 Ti, 4 GB VRAM
Memory	16 GB RAM
Access	SSH and Tailscale
Development	C++ / CUDA experiments on Linux

The hardware gives me a concrete place to explore memory limits, resource sharing, and the difference between an idea and its execution.

Omer Isildak

Mon, 01 Jan 0001 00:00:00 +0000

I’m Omer Isildak. My focus is C++, CUDA, and inference systems—understanding how software uses hardware and how models run as services.

Kernel Nomad is where I keep the technical notes, small experiments, and questions that come out of that work.