We map the ownership and transaction relationships between entities — surfacing circular shareholding, nominee layering, and fabricated revenue that automated screening misses. Verification in 5–10 days, from USD 1,500.
View sample reports — three real cases from the public record →Ownership data across the region is fragmented, multilingual, and walled off by data-localisation laws in Vietnam and Indonesia. Manual Big Four diligence costs USD 50K–500K and takes 4–8 weeks. Automated tools are cheap — and structurally blind to how fraud is actually built here.
eFishery raised over a billion dollars and passed every automated check available to its investors.
The fraud — fabricated revenues routed through connected entities — only surfaces when you trace the relationships between entities, not each one in isolation. That is precisely the gap GroundState is built to close.
Source: Bloomberg, Jan 2025Each of these was adjudicated in court and documented in the public record. Each defeated conventional, entity-by-entity screening. And each left the same structural fingerprint — control and cash moving through connected parties.
These are retrospective illustrations built entirely from public reporting and court records. GroundState did not conduct diligence on these companies; we use them to show the patterns our platform is designed to surface.
Pattern: dual books and revenue routed through ~15 related entities; 400K+ feeders claimed against ~24K real. A relational lens flags the related-party revenue concentration and the asset-count mismatch.
Pattern: ~90% control of a bank held through proxy shareholders, funding thousands of loans to shell borrowers that cycled cash back to the group. This is nominee layering and circular lending — a textbook graph signature.
Pattern: round-tripping through third-party processors made internal cash look like customer revenue; escrow accounts were fabricated. Tracing counterparties and cash direction exposes the loop.
GroundState is a Due Diligence-as-a-Service platform built specifically for Southeast Asia — combining proprietary relational analysis with experienced, on-the-ground regional analysts.
We fuse ownership and transaction graphs to detect circular shareholding, nominee layering, and synthetic revenue flows that text-based tools miss entirely.
A federated architecture trains on jurisdiction-partitioned data — raw records never leave their country, keeping us compliant with Vietnam's and Indonesia's data laws.
Our models read ownership and control from Southeast Asian legal and registry documents across five jurisdictions and multiple languages — where generic AI fails.
When registries disagree — as they routinely do — we reconcile conflicting records into a single, confidence-scored ownership timeline with a clear source trail.
A single, structured process — no black boxes. You see the method and the evidence behind every finding.
We define the target, jurisdiction, and risk questions with you.
Registry filings, media, and financials pulled across five markets.
We build the ownership and transaction graph and run fraud-pattern analysis.
Analysts corroborate flags on the ground — site visits, interviews, documents.
A confidence-scored, evidence-linked verification you can act on.
From rapid AI-assisted pre-screening to full forensic engagement — at a fraction of Big Four cost.
First two engagements available at 30% discount for anchor clients. Contact us to discuss →
Four original technical components addressing structural failures unique to Southeast Asian due diligence.
Uncovers patterns invisible to conventional screening — circular ownership, disguised beneficial control, and fabricated revenue engineered through connected entities. Where standard checks examine companies in isolation, we analyse the web of relationships between them.
Heterogeneous Graph Neural NetworkOperates in full compliance with the data-localisation laws of Vietnam, Indonesia, Malaysia, and the Philippines — without sacrificing regional intelligence. Sensitive records never cross borders, yet the platform keeps learning fraud patterns across the region.
Federated GNNReads and extracts ownership, control, and entity information from regional legal and regulatory documents — across five jurisdictions and multiple languages. Generic AI fails here because it was never built for the region's regulatory vocabulary. GroundState was.
Domain-adapted legal LLMProduces a single, confidence-scored ownership timeline even when government registries disagree. When a tax authority and a business registry record conflicting dates, we surface the most probable truth alongside an auditable record of exactly what each source shows.
Bayesian-HMM reconciliationSoutheast Asia's investors face two inadequate options: Big Four diligence that costs too much and takes too long, or automated tools structurally blind to how fraud actually works here. GroundState replaces both.
| Provider | Structural limitation | GroundState advantage |
|---|---|---|
| Big Four firms | USD 50K–500K per engagement; 4–8 weeks; no AI efficiency; no systematic learning | AI + boots-on-ground hybrid at USD 1,500–30,000; 5–10 days; every engagement improves the model |
| AI background checkers | No physical verification; rules-based; fail on SEA relational fraud patterns | Graph analysis detects circular ownership, nominee layering, and circular invoicing — invisible to text-only tools |
| Centralised data platforms (Refinitiv, LexisNexis) |
Depend on centralised aggregation — legally impermissible in Vietnam and Indonesia | Federated architecture enables cross-border learning without raw data leaving any jurisdiction |
Live in Singapore and Vietnam today, expanding across ASEAN's highest-fraud-risk, hardest-to-verify markets.
GroundState is a Singapore-based, AI-powered due diligence platform for Southeast Asian capital markets, private credit, and MSME lending — a product of Gukom Pte. Ltd., Singapore.
All core IP — model weights, training pipelines, and fraud-pattern datasets — is developed, owned, and domiciled in Singapore.
Engaging with AI Singapore co-innovation programmes, the MAS Fintech Innovation Lab (FSTI), and Enterprise Singapore's Global Innovation Alliance.
Over 20 years across global investment banks and Southeast Asian fintech — FX/rates sales at DBS, BNP Paribas, and Barclays Capital; debt capital markets at Deutsche Bank (second-highest global individual contributor, 2007/2008), Nomura, and Goldman Sachs (Executive Director, SEA institutional clients).
Co-founded Investree — Indonesia's largest SME productive-loan originator — backed by SBI Holdings, MUFG Innovation Partners, and BRI Ventures. Investree was later a direct victim of undetected borrower misconduct: the empirical foundation of GroundState's problem definition.
We're onboarding anchor clients in Singapore and Vietnam at a 30% introductory discount. Tell us about your target and we'll scope the right engagement.
Singapore — GroundState by Gukom Pte. Ltd.
Singapore · Vietnam · Malaysia · Philippines · Indonesia
We scope every engagement within 24 hours
Tell us about your investment and we'll scope the right engagement within 24 hours.