Training Studio: the control panel True Bots trained on

The problem. While training ML agents, we kept going back to the command line and typing long strings of parameter-rich flags. It was really slowing us down, but it was all programmatic.

What I built. A GUI that works directly with the YAML configs, and a way to drive the launch controls that is more intuitive, repeatable and, most importantly, automatable. It came to 21,000 lines that I hand-rolled piece by piece:

a live, interactive interface for the YAML training configs;
a script that extracted the validation rules from ML-Agents itself, so they followed new versions instead of drifting;
a validator that checked the machine and the training environment against them: hardware, platform, paths and the ML components;
an environment builder that built the right environment when one wasn't there;
run launch, a queue, and live monitoring.

Training Studio: the config editor and launcher, beside the environment validation, all passing

What happened. Our core team of three trained on it from 2024 into 2025.

What I learned. Build for the automation you can see coming. I gave Training Studio command-line and API entry points from the start, because we expected LLMs and agents to drive training before that was a common thing.