Uncover hidden liabilities, mistranslations, and accounting anomalies in Chinese equities. Deterministic, institutional-grade analysis.
Backtested against the Country Garden (2007.HK) 2022 Annual Report. Our engine generated a HIGH RISK signal six months before the $11B default.
RED FLAG — GOING CONCERN DISCLOSURE: Management included a specific "Going Concern" assessment relying on RMB 266 billion in "unused loan facilities" and RMB 20 billion in new medium-term note quotas to survive the next 12 months [Page 135]. RED FLAG — RELATED PARTY BAILOUT: Controlling shareholder provided an interest-free, unsecured loan of HKD 5.055 billion (~RMB 4,500 million) with a 37-month maturity [Page 245].
Earnings backed by hard cash flow (±10% tolerance).
Financial Statement Cross-Check (P&L ↔ Cash Flow)
Variance: 0% · Different line items — expected
Fiscal Year Alignment
No fiscal year specification discrepancies detected.
Restatement / Reclassification Scan · 1 flag(s)
1 footnote flag(s) detected (0 HIGH, 0 MEDIUM)
Trigger: "adopted HKFRS 17"
Four localized forensic modules — each targeting a specific vector of financial manipulation in Chinese equities.
Cross-examines English and Chinese filings across 10 hunt categories to surface omissions, softened disclosures, and numerical deltas hidden in translation — the "Ghost Cards" that only bilingual analysis can reveal.
8-variable probabilistic model for earnings manipulation detection. Calibrated against the -1.78 institutional threshold with full variable-level decomposition — DSRI, GMI, AQI, SGI, DEPI, SGAI, LVGI, and TATA.
Measures the gap between reported earnings and operating cash flows. A semicircular gauge visualizes the safety range (±10%), instantly flagging entities with aggressive accrual-based earnings.
Scans for inventory build-up signals, depreciation policy changes, and marketing-to-revenue efficiency anomalies. Each finding is tagged with severity and deep-linked to the source PDF page.
A deterministic, math-backed pipeline that transforms raw financial filings into actionable forensic intelligence — enforced by a 9-node agentic swarm and rigid Pydantic schemas.
Financial language processing driven by specialized multi-agent architectures that cross-examine corporate disclosures independently.
Ingestion
A distributed swarm utilizing dual-speed logic to route, translate, and verify granular numerical claims at scale without bottlenecking.
Processing
Eliminates LLM math hallucinations. We force extracted values into rigid Pydantic schemas before routing them through deterministic formulas.
OutputDeploying agentic intelligence across primary Chinese and Hong Kong equity markets.
Institutional-grade forensic intelligence built by a team with decades of on-the-ground China operations, cross-border scaling, and AI engineering.
Request early access to the Vestigia Labs institutional terminal. Spots are limited — currently supporting HKEX-listed entities with bilingual EN/ZH coverage.