"Trained an LLM": Pre-Training vs. Fine-Tuning vs. Calling an API
New'Trained a model' or 'built an LLM' on a resume can mean three jobs that differ by orders of magnitude in skill and cost. Pre-training builds a model from scratch on huge amounts of text using thousands of GPUs — only a handful of companies do it. Fine-tuning adjusts an existing model with a smaller, specialised dataset. Calling an API means using someone else's model, unchanged, inside an application. All three are legitimate; only the first two involve training.
How to tell them apart on a resume
Pre-training
Thousands of GPUs, distributed training, tokens in the trillions, training stability, data mixtures — at an AI lab or a large tech company's research group.
Fine-tuning
LoRA, QLoRA, Hugging Face, a curated dataset of thousands of examples, evaluation before and after, a single GPU server or a cloud fine-tuning service.
Calling an API
OpenAI, Anthropic or Gemini APIs, prompts, RAG, LangChain, building a product feature on a model someone else trained.
The question that settles it
“When you say you trained the model, did you start from scratch, adjust an existing model with your own data, or use a model through an API — and how many GPUs were involved?”
Read the full definitions
Open the full tool for the other look-alike pairs, role profiles, and the JD decoder.