Regime Alpha

The problem

Full-period averages can hide meaningful changes in how a portfolio behaves. Holdings that appear diversified over an entire sample may become more closely related in particular conditions. Regime Alpha makes those differences available for inspection.

What I’m building

The platform connects market and macroeconomic inputs with portfolio factor decomposition and regime-conditioned diagnostics. The research includes mixture models, hidden Markov models, stochastic state transitions, and the interpretation of regime probabilities.

• Inspect changes in volatility, correlation, and drawdowns.
• Understand factor concentration and portfolio risk contributions.
• Compare relationships across market states.
• Keep the data, assumptions, and saved analysis connected.

From inputs to a review

The workflow begins with portfolio holdings and selected data. Explicit model-building steps produce saved analyses, allowing edited inputs to remain separate from a completed result. Data quality and readiness checks are intended to prevent stale or invalid inputs from entering a build.

My work includes modeling, Python data pipelines, analytical workflow design, and review of AI-assisted code changes. The aim is an inspectable research process with clear boundaries around what each result means.

Interpreting the output

The outputs are historical, model-dependent diagnostics. They do not establish causality, calibrated forecasts, or investment outperformance. The public portfolio mechanics example uses synthetic inputs and is labeled accordingly; it is not an empirical performance record.

The public website explains the platform and its methods. Application access is approval-based.