← Back to Journal

Practicing Working with Hugging Face

SEPTEMBER 13, 2026 · Career

I finished DataCamp's Working with Hugging Face course, the third course on the way to the Associate AI Engineer for Data Scientists track. Hugging Face itself wasn't new to me, I'd used it in my deep learning course and in a course on training your own LLM in grad school. Going through the Hub, pipelines, and AutoClasses as their own subject, instead of as one piece of a bigger assignment, was.

Working with Hugging Face course completion badge

I knew how to use it, not what was happening underneath

In grad school I'd load a model, fine-tune it, and move on to the next step of whatever the assignment was. I never had a reason to stop and ask what AutoModel and AutoTokenizer are actually picking for you behind that generic name, or when you'd want a raw model and tokenizer instead of a pipeline. Now I do.

Searching the Hub is its own skill

Filtering models by task, author, or popularity sounds simple until you're looking at a few hundred checkpoints for the same task with no clear sense of which one fits your data or your compute budget. Before this, an instructor or a project spec had always pointed me at a specific model. This was the first time I had to narrow it down myself.

What's next

More of the track. Whatever comes next is probably further from what I already touched in grad school, so I'll actually have to pay attention.

Career