Command line script and Python library to work with time-series data.

Given two 2D arrays (grids), perform a windowed linear regression between the two. Aim: removal of correlated component between two variables.

A collection of substantial projects I completed for the CS156 Machine Learning course.

Simulates a rolling window strategy that continuously optimizes and deploys the bot with updated thresholds. This project serves as the backtester for the original strategy.

End-to-End Python replication of Wu et al.'s (2026) Sentiment Networks methodology. Quantifies directed sentiment contagion across 34 tech equities via Transfer Entropy, Markov bootstrap significance filtering, weighted PageRank centrality, and Chu-Liu/Edmonds Maximum Spanning Arborescence. Compares News vs. Social Media information topologies.

Empirical forecasting of the equity risk premium using historical financial predictors and time-series regression methods.

SQLite SQL practice project covering daily metrics, date spine, running totals, rolling windows, and late-arriving data patterns (GROUP BY + window functions).

Operational monitor for multi-agent AI. Per-action observations, per-agent rolling trends, baseline-relative alarms, named recommendations.

Sliding window rate limiter implemented in Golang/ Go using redis

Plan low-cost primer automations for Codex and Claude Code rolling quota windows.

End-to-End Python implementation of "Reflexivity as Prompt" (Park, 2026). Evaluates LLMs as financial forecasters during boom-bust cycles using George Soros's Reflexivity Theory as a zero-shot prompt scaffold. Implements P/E-ratio-invariant normalization, buy-and-hold calibration gates, and a structured diagnostic reproducibility audit taxonomy.