teacher-support-studio

Teacher Support Studio is an end-to-end educational machine learning project that predicts a student's probability of first-attempt success on upcoming skill practice and turns those estimates into teacher-facing, evidence-grounded guidance. It combines leakage-safe longitudinal feature engineering, interpretable and nonlinear classification models, SHAP explanations, and a deployable FastAPI application. The modeling uses the ASSISTments 2009–2010 Skill Builder dataset, with 259,386 filtered interactions from 4,163 students and 170 documented predictors constructed only from information available before each prediction. Models include L2-regularized logistic regression and tuned XGBoost; the best test ROC-AUC is 0.7000 with XGBoost. The application lets a teacher select a synthetic class and student, review estimated readiness across named skills, and ask questions grounded in the displayed evidence. For each skill, it reconstructs the student's state from all prior interactions, creates up to ten plausible next-practice contexts, and batch-scores those scenarios with the persisted XGBoost model. A LangGraph workflow gives OpenAI only the selected metrics and supporting evidence for a structured response; if OpenAI is unavailable, the endpoint returns deterministic local guidance.