Customizing an Open-Source Model Without an ML Team

Sun, 05 Apr 2026 00:00:00 +0000

We all consume AI models through APIs. That works, but they don’t always behave how we want, and they’re often less efficient than they could be for the specific tasks we have.

I’ve always assumed that doing something about that, actually customizing a model, required a specialized ML team and serious GPU infrastructure. Is that still true in 2026?

I wanted to find out. When Google released Gemma 4, I set up a hands-on survey: how far can one engineer get with open-source tools, local hardware, and AI agents helping navigate the process?

Model-Training on Charlie Hulcher

Customizing an Open-Source Model Without an ML Team