About · author and research program
Conrad Gann
Quantitative researcher and builder with a rare mix of macro forecasting, time-series modeling, and equity statistical-arbitrage experience — now focused on making portfolio risk decomposition operational: variance attribution, hedge ratios, subsector exposures, and model-governed reporting for allocators and portfolio managers.
Track record
Blue Water MacroNew York / Larkspur, CA
2021–present
Founder. Blue Water Macro Corp was launched on May 28, 2024 as a financial-technology company, building and operating the RiskModels engine — the relaunched Cerebellum risk model, now a full production environment on EODHD data (held under a redistribution license), on Dagster, Python, SQL, Zarr, and GCP. The earlier Blue Water Macro LLC is no longer a registered investment adviser. Macro-to-market dataset consulting for Hull Tactical (manager of the HTUS ETF).
Cerebellum CapitalSan Francisco, CA
2010–2021
Led an 8-person statistical-arbitrage research team building strategies with genetic-programming / symbolic learning (DEAP) over stat-arb terminal functions; modeled SPY with PCA + regression on GICS sector returns; used Gaussian Mixture Models to form correlation groups; and built and traded an APAC/China stat-arb book with Myriad Asset Management.
TrimTabs Investment ResearchSausalito, CA
2005–2009
Built a liquidity-based GICS sector-rotation model funded with a $150M allocation from Fortress's Drawbridge fund, and a 2008 SPY timing model from 21 flow- and sentiment-related datasets that ran for 14 years with few changes. Expanded coverage to Europe and Asia, sector liquidity, ETF flows, and hedge-fund flows.
Citadel Investment GroupChicago, IL / San Francisco, CA
2002–2003
Frequency and severity distributions for industry-loss-warranty (ILW) property-catastrophe reinsurance models.
Putnam CompaniesBoston, MA
1992
Global Fixed Income intern — modeled optimal currency hedging for a global bond portfolio.
Federal Reserve Bank of San FranciscoSan Francisco, CA
1989–1991
Research coordinator in macroeconomics, producing forecasts for the FOMC briefing book. Built a Bayesian-prior vector autoregression (GDP, CPI/PPI, capacity utilization, payrolls, unemployment, FX, S&P 500, oil, and trade) with a Kalman filter for forecasting on partially-updated data, and provided MATLAB support to Tom Sargent — later a Nobel laureate in Economics — on rational-expectations equilibria.
Research focus
Measurement
Replace broad labels with decomposed variance shares and factor lineage.
Execution
Translate orthogonalized betas into ETF hedge ratios and reproducible workflows.
Governance
Keep model outputs auditable, explainable, and tied to explicit data windows.
RiskModels.org
RiskModels.org is the public research and methodology layer for the RiskModels ecosystem. It hosts the primer, research articles, reference map, and educational decomposition workspace. Product surfaces for API access, SDKs, CLI workflows, and model operations are intentionally separated at riskmodels.app. Together they form a quantitative truth layer for AI-native finance: read the method here, call the structured, time-stamped model there — the same decomposition, whether a person reads it or an agent cites it.
The Analyst Workspace — where the decomposition runs on a whole portfolio — is in active development. You can explore the live demo today; running it on your own holdings is invite-managed while it matures.
Institutional posture
The writing style is deliberately narrow: define the measurement problem, show the math, state the edge cases, cite the lineage, and point to reproducible tooling. No performance claims, no black-box alpha narrative, and no marketing claims where a decomposition table would be more precise.
Contact: support@riskmodels.app
Read the literature map →