Fine-tuning a small model to judge CVs against job postings

A recruitment company wanted an assistant that writes job postings, suggests interview questions and scores how well a CV fits a role. Calling a large model for every request costs more as volume grows.

I built a pipeline that generated and checked training data, then fine-tuned a small model to do these jobs at lower cost.

The client got a path to run these tasks at lower cost than a large model.

• GPT-4o
• LoRA fine-tuning
• unsloth
• Python