Adaptive Testing Framework for AI Models (Psychometrics in AI Evaluation)

Adapting Questions to the student ability

Score Impact Analyzer is a data-driven tool that simulates Digital SAT adaptive scoring to identify which missed questions have the highest impact on a student’s final score, helping optimize study focus for maximum improvement.

This project considers the use of p-value weighting to voxel-based lesion behavior mapping (VLSM) studies. Our methods are demonstrated on an aphasia study investigating which regions of the brain are associated with the severity of language impairment among stroke survivors.

Item Response Theory (IRT) — the framework behind adaptive tests like GRE and PTE. Designed for AI engineers and data scientists to grasp core IRT concepts for adaptive testing systems.

AI-powered adaptive testing engine using FastAPI, MongoDB, and IRT-based difficulty adjustment.

An adaptive testing backend built with FastAPI and MongoDB that dynamically adjusts question difficulty based on student performance and generates personalized study recommendations.

Browser-based auditory temporal processing training app. Research-grade psychoacoustic exercises (gap detection, TOJ, dichotic TOJ, duration reproduction, pitch discrimination, pattern detection) with adaptive staircase procedures and sample-accurate Web Audio API. Methodology follows Fostick & Babkoff (2022).

Adaptive GRE prep tool using 1PL IRT with Fisher Information item selection for real-time ability estimation. LLM generates personalized study plans with curated resources. Built with FastAPI, MongoDB Atlas, Clerk Auth, and OpenRouter. 74 tests, Docker support, structured logging.

IRT-based adaptive testing engine with real-time question selection

1D Adaptive Testing Engine using Item Response Theory (IRT) | FastAPI · MongoDB Atlas · GPT-4o-mini | Dynamically adjusts GRE question difficulty based on real-time ability estimation

1D adaptive diagnostic engine, uses item response theory (Rasch Model) to dynamically select GRE - style questions based on student ability then generates a pesonalized AI study plan.