The wrangler

LAST SEEN: at a terminal, mid-finetune

You don't need a department. You need one person who has actually done the thing you're trying to do.

I train and finetune generative models — video, audio and image — and build the automated pipelines that put them to work. The evidence is public: open LoRAs on Hugging Face, and a content pipeline that has published daily, unattended, since the day it was let loose.

I ride open-source first and manage my own inference - predominantly locally by choice — because owning your weights and your iron is cheaper, faster and more private than renting someone else's. When a closed model genuinely wins, I'll say so and use it.

One wrangler, no handoffs: the person you talk to is the person who preps the data, trains the model and hands you the reins.