Mini-project
Banking & Fintech
Fine-tuned complaint classifier
Fine-tune a small model to categorise complaints, and compare it with prompting a large one.
The brief
A bank wants complaint sorting it can run on its own servers, without sending customer text to an outside API. Fine-tune a small open model to sort complaints into categories, and compare it with prompting a large hosted model on accuracy, speed and cost. Good looks like: a clean training set, a fair test, and a recommendation the IT security team would accept.
What you will build
- Prepare labelled complaints and a held-out test set
- Fine-tune a small open model with LoRA
- Prompt a large model on the same test set
- Compare accuracy per category, latency and cost
- Recommend one, with the privacy trade-offs
Free, open data to build it with
- Consumer Complaint Database ↗ Consumer Financial Protection Bureau
Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.