Travel document / Résumé

Akshay Byagathvalli

Data Science, Artificial Intelligence, and Research

Profile

I investigate complex questions through data, build intelligent systems, and communicate results through rigorous analysis and visualization.

Selected research

World Cup Player Impact

An end-to-end study of 2,927 player-tournament observations that combines role-relative measurement, reliability adjustment, nested validation, temporal testing, clustering, and explainable machine learning.

Read the project

Pressure, Production, and Noise

A reproducible study of whether normal-condition offensive traits explain or predict the preservation of scoring efficiency during NBA clutch situations, with explicit non-clutch baselines, repeated validation, shrinkage, simulation, and multi-season stability checks.

Read the project

Concentration, Chokepoints, and Supply-at-Risk

A reproducible study of geographic concentration, country-mineral networks, structural vulnerability, and simulated supply disruption across a synthetic panel of 24 critical minerals and 35 producer countries.

Read the project

The Economics of LEGO

A reproducible 50-year study of 18,540 LEGO set records joined to original US retail prices, combining hedonic and quantile regression, mixed models, structural-break tests, theme diversification indices and twelve machine-learning models with controlled ablations.

Read the project

Armed Societies, Unequal Outcomes

A reproducible cross-national study linking Small Arms Survey civilian firearm estimates to UNODC firearm homicide rates for 96 countries, with twelve regression specifications, fourteen nested-cross-validated learners, feature ablations, spatial autocorrelation tests and a 34-specification robustness matrix.

Read the project

Research approach

Questions → evidence → useful tools

Frame testable questions, document data provenance, compare methods honestly, and communicate findings through accessible visual explanations.

Portfolio field guide

Technical toolkit

Methods used in published work

Pythonpandasscikit-learnOptunaSHAPPlotlystatsmodelsnba_apiMonte CarloNetworkXLightGBMXGBoostShapely