Built this: Explainable Student Performance Prediction Platform
An end-to-end ML platform that validates student data, benchmarks nine model types from logistic regression to an LSTM, tracks runs in MLflow, explains predictions with SHAP and LIME, and serves them through FastAPI and Streamlit.
On 28 July 2026 I committed an explainable student-performance platform. It reads almost like a summary of my whole ML phase in one repo.
Pipeline
From the README’s architecture diagram:
- Data: UCI student dataset or a synthetic raw CSV → Pandera validation and cleaning → feature engineering with a
ColumnTransformer→ stratified train/validation/test splits. - Model benchmark: logistic regression, decision tree, random forest, XGBoost, CatBoost, LightGBM, a PyTorch MLP, a PyTorch LSTM for temporal data, and a FLAML AutoML baseline.
- Tracking: every run logged to MLflow.
- Explainability: SHAP and LIME explanations.
- Serving: a FastAPI REST API (port 8000) and a Streamlit dashboard (port 8501).
Where it connects
It reuses the ideas from my scikit-learn pipelines, cross-validation and boosting notes, and adds the explainability and serving layers those posts didn’t cover.
#machine-learning#explainability#mlops
Sources
- SahajIVVIX-1/student-performance-platform (GitHub) — GitHub
- First commit 37f2457 — GitHub
- SHAP documentation — SHAP
This is a summary written from the linked public sources. Figures and numbers are as reported there.