Projects
Curated work organized by maturity — from flagship systems to archival records.
How can forecasts, scenarios and allocation rules be unified into a single auditable decision architecture?
End-to-end framework linking econometric forecasts, scenario simulation and allocation rules for operational decision-making.
Can local interaction rules reproduce macro-level market regimes without imposing equilibrium assumptions?
Agent-based environment for studying how micro-level behavior aggregates into macro patterns relevant to systemic risk analysis.
Can hierarchical models deliver stable default estimates when panel data suffers informative missingness?
Hierarchical models applied to longitudinal credit data with emphasis on interpretability, calibration and out-of-sample stability.
What infrastructure guarantees reproducible, versioned features across forecasting and risk models in production?
Data engineering architecture for versioning, validation and serving features used across forecasting and risk models.
How should posterior uncertainty in structural identification inform macro-financial policy analysis?
Bayesian estimation of structural VAR models to quantify shock propagation and policy-relevant impulse responses.
Formalized hypotheses before estimation reduce specification drift in applied research.
Exploratory tooling to formalize economic hypotheses before estimation, preserving statistical rigor in the modeling loop.
Echo-state dynamics capture short-horizon variance structure missed by linear GARCH specifications.
Experimental comparison of echo-state networks against classical GARCH-family models on high-frequency return series.
Constraint-based structure learning can surface stable causal graphs under regime shifts in panel data.
Pilot study applying PC-stable and score-based algorithms to equity factor panels with rolling validation windows.
Depth imbalance at short horizons contains information orthogonal to mid-price momentum signals.
Feature engineering sandbox on L2 snapshots with leakage controls and transaction cost-aware evaluation.
Can hierarchical models deliver stable default estimates when panel data suffers informative missingness?
Hierarchical models applied to longitudinal credit data with emphasis on interpretability, calibration and out-of-sample stability.