Argus

Summary

Argus integrates 134 quantitative risk modules across six analytical layers (banking supervision, systemic stability, macroeconomic conditions, cross-border flows, integration, and crosscutting themes) into a single composite risk score. It is backed by 13 data collectors, 28 AI-powered analyst tools, 10 CCAR-style stress scenarios, and publication-quality briefing generation, running on 1.3 million rows of real FDIC, FRED, Yahoo Finance, OFAC, and FATF data. The platform supports bank examiners, risk officers, macro-prudential researchers, and portfolio managers who need transparent, reproducible analytics.

What it is

Argus is research infrastructure that Deluair Consultancy operates and makes available to clients through advisory engagements. It is an open, fully reproducible macro-financial risk observatory built around 284 Python modules and 3,766 tests, with zero lint errors. Five core risk layers (banking, systemic, macro, cross-border, crosscutting) plus an integration engine produce a single composite score, calibrated against historical episodes including the Global Financial Crisis, the Euro Crisis, the Taper Tantrum, COVID, and the SVB failure. Bank-level supervisory scoring uses a CAMELS-calibrated MSPS framework; systemic risk is measured through a full suite of CoVaR, SRISK, and MES estimators; macroeconomic conditions are summarized through a 12-component PCA-based Financial Conditions Index, credit impulse, and BVAR forecasting.

For clients, Argus replaces the closed-vendor stack (Bloomberg Terminal, BlackRock Aladdin, internal supervisory tools) with a transparent, auditable platform that they can interrogate, extend, and validate. Engagements typically combine the standing analytical layers with custom stress scenarios, bespoke briefings, and AI analyst workflows that span 28 structured tools across bank health, systemic risk, macro conditions, cross-border flows, and contagion simulation. A 15-agent specialist team (CRO orchestrator plus 14 domain analysts) coordinates morning briefings, supervisory memos, and thematic research. Outputs are delivered as publication-quality HTML and PDF briefings with Plotly visualizations, alongside the underlying SQLite warehouse and source code, so clients retain full methodology transparency rather than a black-box score.

Methodology

• Six-layer risk architecture with 134 modules: 24 banking, 33 systemic, 35 macro, 27 cross-border, 10 integration, and 5 crosscutting themes (climate, cyber, geopolitical, fintech, pandemic).
• Bank-level supervisory scoring (MSPS) on a CAMELS-calibrated 1 to 5 scale, with panel logit distress early warning and peer percentile ranking across 4 FDIC peer groups.
• Full systemic risk suite: CoVaR, SRISK (DCC-GARCH with MLE), MES, absorption ratio (PCA), Eisenberg-Noe network clearing with RAS, Diebold-Yilmaz connectedness, and Kritzman-Li turbulence.
• Macro analytics: 12-component PCA Financial Conditions Index, credit-to-GDP gap (one-sided HP filter, BIS lambda), Nelson-Siegel term premium, Hamilton regime switching, and BVAR forecasting with a Minnesota prior.
• Cross-border AML and TBML analytics combining misinvoicing detection, unit-price z-scores, FATF technical and effectiveness composites, sanctions network analysis, and CFaR via quantile regression.
• CCAR-style stress testing across 10 built-in scenarios (severe recession, stagflation, sovereign crisis, dollar crisis, pandemic shock, cyberattack, China hard landing, rate shock up and down, geopolitical escalation), plus a custom shock-vector builder with cross-layer propagation.
• Integration engine: weighted composite (0.25 per core layer), sensitivity analysis, attribution decomposition, ensemble scoring, Granger causality network, and crisis backtesting against GFC, Euro Crisis, Taper Tantrum, COVID, and SVB.
• AI analyst with 28 structured tools and a plan, execute, synthesize research mode, plus a 15-agent team (CRO orchestrator and 14 domain specialists) for autonomous briefing generation.

Data sources

• FDIC BankFind API and Call Reports (4,313 US insured institutions, 58 metrics per bank, 2024Q1 to 2025Q4)
• Federal Reserve Economic Data (FRED): 25 macro and financial series including VIX, spreads, yields, CPI, and SLOOS, 2000 to 2026
• Yahoo Finance equity returns: 27 BHC tickers plus KBE, daily 2020 to 2026
• OFAC Specially Designated Nationals list (full SDN extract)
• FATF mutual evaluation ratings: 40 jurisdictions, 51 compliance items each
• BIS (SDMX), BACI / CEPII, and UN Comtrade for cross-border trade and capital flows
• FinCEN / BSA filings, CPMI payments statistics, and TIC / Treasury international capital data
• FDIC bank failures dataset (3,626 US bank failures, 1985 to 2011)

Deliverables when used in engagements

• Composite macro-financial risk score with full layer-by-layer attribution and a signal level (NORMAL, WATCH, ELEVATED, HIGH, CRITICAL).
• Bank-level MSPS supervisory scorecards with peer percentile ranking, distress probability, and CRE, HTM, deposit-flight, and wholesale funding diagnostics.
• CCAR-style stress test results across 10 standard scenarios plus client-defined custom shocks with cross-layer propagation.
• Publication-quality briefings (stability assessment, flash alert, supervisory memo, thematic brief, morning digest) delivered as HTML with Plotly charts and PDF export.
• Reproducible analytical environment: SQLite warehouse with 1.3 million rows of public data, 284 Python modules, and 3,766 tests transferred to the client.