DATA SCIENCE
DataSense AI — AutoML & Explainability Platform
End-to-end data science web app: upload a CSV and get quality scoring, interactive EDA, a 10-model AutoML leaderboard, SHAP feature importance, anomaly detection, and a PDF report — all in one pipeline.
Public demo
10 — AutoML models benchmarked
The problem
Data scientists repeat the same boilerplate (profiling → cleaning → model training → SHAP) on every new dataset, burning hours before getting to the actual insight work.
The approach
FastAPI backend orchestrates Pandas profiling, Isolation Forest anomaly detection, and a 10-model AutoML grid (XGBoost, LightGBM, sklearn). SHAP TreeExplainer explains the winning model. Claude API powers a contextual AI analyst chat that understands the uploaded dataset.
The results
10-model leaderboard with live progress tracking. SHAP waterfall, summary, and dependence plots generated automatically. Claude-powered chat answers dataset-specific questions with full context. PDF reports exported via FPDF2.
Stack: React 18, Vite 7, FastAPI, XGBoost, LightGBM, SHAP, Claude API