Data

24-hour optimization sprint on a supply-chain case blending MIP, clustering, and time-series signals.

Causal inference on subsidy batches to estimate true incremental GMV and reallocate budget to higher-ROI segments.

Empirical valuation lens linking innovation-heavy productive forces to market repricing with panel-style evidence.

CV pipeline from HOG/KNN baselines toward CNN/ResNet-style models with explainability hooks.

A multi-asset portfolio strategy balancing alpha generation, transaction costs, and risk management across equities, fixed income, commodities, FX, and cash.

Clustered thousands of AI-creator uploads to find better time-slot × scale combinations and quantify view lift.

Hazard exposure scoring with socio-economic layers and parametric insurance recommendations.

Cleaned Likert survey data on study habits and expectations with clear reporting-ready visuals.

JSON→tabular pipeline with dialogue metrics for rhythm, diversity, and engagement in role-play training data.

Univariate and multivariate forecasting experiments to relate AI-cycle narratives to NVDA price dynamics.

Compared regression and nonlinear forecast specs to test robustness and omitted-variable bias on earnings drivers.

Red vs white Vino Verde modeling with selection, stability checks, and interpretable drivers of perceived quality.

Explored regional employer–income patterns with tree ensembles, KNN, NB, and regression baselines.

Composite accessibility index with normalization, weighting, and defensible ranking across SA2 units.

Accuracy vs runtime trade-offs across classical learners with stratified CV on medical vs sensor data.

Large-scale comment mining to trace sentiment shifts, persuasion routes, and herding in a major pop-culture event.

Predictive modeling for quick-service daily sales with elastic net, ridge/lasso, KNN, and rigorous CV.

My APMCM paper: optimization-first model, parameter sensitivity, and a tight writeup for our problem track.

Our MCM/ICM entry: clear assumptions, statistical core models, and evidence-backed answers for the chosen problem.