Ugo Chukwu

Your first fine-tune should teach you the whole loop.

I walk through it with Unsloth: data → baseline → training → evaluation → local model → Python.

Including a mistake I made—and why a completed training run doesn't mean you have a better model.

Full tutorial below 👇

https://youtu.be/YxXBojdD1as

I Fine-Tuned a Model. The Failures Taught Me More Than the Winning Score. What a classification win over GPT-4.1—and failures on unfamiliar regulatory questions—taught me about building useful domain intelligence.

Check it out

Just published: Multimodal Video Pipeline.

Generating video is one thing. Building a system that can reliably manage multiple scenes, models, failures, retries, assembly, and exact-version approval is another.

I built a production-style pipeline around:

Storyboard → Scene Routing →...

Just published: Static Creative Studio.

Generating an ad image is easy. Keeping it aligned with approved strategy, claims, copy, feedback, and reviewer authority is harder.

I built a governed creative workflow around:

Strategy → Generation → Review → Versioning → Approval →...