back to projects
xai-gpt-shap
xai-gpt-shap
Python library that combines SHAP analysis with GPT to generate natural-language explanations of ML model predictions — tailored to different audiences, from beginners to researchers.
2024Personal / Master's
PythonSHAPOpenAIXAIscikit-learnpersonal
// purpose
Thesis work on explainable AI — making ML predictions interpretable through LLM-generated explanations.
// takeaway
LLMs are good at translating technical output into plain language — the challenge is keeping them grounded in the actual data.
// stack
PythonSHAPOpenAI APIXGBoostscikit-learnPoetry
// challenges
01
Generating explanations that are accurate to the SHAP values, not hallucinated.
02
Supporting multiple model formats (ONNX, Pickle) without coupling the library to a specific framework.