Mini-project
Automotive, EV & Aerospace
Fine-tuned fault code explainer
Fine-tune a small model to explain diagnostic trouble codes, and compare it with a large model.
The brief
Mechanics and owners look up diagnostic trouble codes (DTCs) one by one. Fine-tune a small model that explains each code, its likely causes and how urgent it is, and compare it with a large hosted model on accuracy and made-up answers. Good looks like: a training set from a trusted code list, a test set, and a careful check for confident wrong answers.
What you will build
- Build code, meaning and cause pairs from a public code list
- Fine-tune a small model with LoRA
- Test both models on held-out codes
- Measure accuracy and made-up answers
- Recommend one, with its limits
Free, open data to build it with
- OBD-II trouble codes ↗ GitHub
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.