Developed and contributed 12 applied mathematics, economics, and geology questions to Humanity's Last Exam (HLE) benchmark dataset. HLE aims to aggregate some of the toughest academic problems to test LLM's technical reasoning in the age of CoT.

Graduate research study on improving real-world reasoning capabilities in Large Language Models (LLMs) for robotic control applications. Exploring Reinforcement Learning (RL), multi-modal inputs, and fine-tuning techniques to enhance LLM-based robotic decision-making in dynamic environments.

Developing a framework for ETF-only portfolio construction, applying Hierarchical Risk Parity (HRP) to optimize asset allocation based on predetermined sector allocation and risk tolerance.

Conducted a comprehensive analysis of key enrollment growth factors within all US Student Housing real estate markets. Developed models for factor identification, applied Principal Component Analysis (PCA), Graphical Lasso, and Boosted Random Forests.

This article explores the integration of reinforcement learning techniques in robotic control using LLMs as instruction/plan generators.It covers policy optimization (PPO vs. GPRO), training processes and logistics, and performance.

This article explores deriving option-implied probability distributions from butterfly spreads to evaluate market expectations. It covers risk-neutral pricing, volatility surface interpretation, distribution generation from market prices of options with different strike prices.