A comprehensive benchmarking framework for Simulation-Based Inference algorithms. Compares BayesFlow, SNL, and Affine Flow Matching on synthetic and real-world tasks, with a complete evaluation pipeline including posterior calibration and recovery metrics.

Dynamic self-rewarding framework for aligning medical LLMs. Uses a two-tier GPT-4o judge system with multi-round DPO to achieve alignment without human-annotated supervision, tested on HealthCareMagic, PubMedQA, and MedMCQA.

Contributed to Texera, a collaborative ML-based data analysis workflow platform at UC Irvine. Integrated AI-driven pipeline automation, built automated report generation, and enhanced data cleaning and visualization capabilities.

Mathematical model of bacterial infections and antibiotic treatments using systems of ODEs. Leveraged Fourier analysis and Laplace transforms for resistance prediction, and optimized drug schedules through numerical simulation.

Analyzed economic uncertainty during the pandemic using stochastic models, Monte Carlo simulations, and Bayesian inference techniques. Built computational tools for uncertainty visualization in economic forecasting.

Built a population dynamics model incorporating age-stratified fertility rate adjustments. Performed parameter sensitivity analysis for demographic forecasting optimization.