Connect public data, corporate intelligence and evidence-first AI in one investigation workspace for banks, NBFCs, fintechs and risk teams.
Counts describe the designed platform scope, not performance claims.
User roles
Report types
Specialised AI agents
Approval stages
File formats supported
Webhook events
Registries, documents, websites and news live in separate tools. Shared directors, conflicting addresses and stale records surface late, or not at all.
Corporate, regulatory, legal, news and domain data in different formats.
“Pvt Ltd”, “Private Limited” and spelling variants hide the same entity.
Shared directors, addresses and contact channels rarely show in a flat review.
Scores without evidence cannot stand up to audit.
Enter a company, CIN, GSTIN, person, phone, email or website. Connectors query permitted sources in parallel.
Fuzzy, phonetic and alias matching link records with a visible confidence level.
A time-aware graph shows who is linked to whom, and when.
OCR extracts fields from uploaded documents and compares them with public records.
Signals and contradictions arrive with evidence. Analysts confirm, dismiss or escalate.
Analyst, manager and compliance review. Versioned, QR-verifiable output.
Six working demos that run in your browser on fictional data. The logic you see is real; the data is demo data.
Drag nodes to rearrange. Move the year slider to see relationships appear and end.
Fictional entities. Relationship dates are illustrative.
Try typos, “Pvt” vs “Private”, or word order. Scores use normalisation, edit distance and token overlap on a fictional dataset.
6 of 9 demo borrowers match → create review signal
| Borrower | Directors shared | Result |
|---|---|---|
| ApexNova Technologies Pvt Ltd | 3 | Signal: create review signal |
| Nova Retail Pvt Ltd | 3 | Signal: create review signal |
| Apex Logistics LLP | 2 | Signal: create review signal |
| Meridian Traders Pvt Ltd | 3 | Signal: create review signal |
| ApexNova Services Pvt Ltd | 1 | No signal |
| Harbor Foods Ltd | 2 | Signal: create review signal |
| BlueRiver Infra Pvt Ltd | 2 | Signal: create review signal |
| Apex Nova Technology Solutions | 0 | No signal |
| Orion Textiles Private Limited | 0 | No signal |
Rules are evaluated live against the demo borrowers. In the product, rules are versioned and can be dry-run on historical cases.
| Field | Document | Registry (demo) | Match |
|---|---|---|---|
| Company name | ApexNova Technologies Pvt Ltd | ApexNova Technologies Private Limited | Strong · 100% |
| Address | Pune | Pune | Strong · 100% |
| Director | A. Mehta | Arjun Mehta | Probable · 78% |
| Incorporation year | 2019 | 2019 | Strong · 100% |
Fields are consistent with the demo registry record.
Extracted fields are pre-set sample values standing in for OCR output. Comparisons against the demo registry record are computed live.
4 sources show different business addresses. Demo data. Evidence, sources and confidence appear here in the product.
Arjun Mehta is a director of 4 demo companies. Demo data. Evidence, sources and confidence appear here in the product.
Incorporation year differs between certificate and registry. Demo data. Evidence, sources and confidence appear here in the product.
3 related companies were created within 3 years. Demo data. Evidence, sources and confidence appear here in the product.
A 2024 notice is linked to a related entity. Demo data. Evidence, sources and confidence appear here in the product.
5 of 5 sources queried
Full demo coverage. Coverage is an availability measure, not a credit score.
VER-2026-000184 · v1 · draft · demo
The analyst experience, from portfolio view to a verifiable report. Demo data only.
Three records contain materially different addresses. Requires analyst review.
ConfirmDismissFalse positiveEscalateNo report content is exposed publicly.
Each layer sits behind an interface, so sources, search, graph and AI providers can be swapped without rebuilding. Tap a layer.
Fuzzy, phonetic and alias matching link spellings and sources, and always show how sure the match is. A match is never a confirmed identity until an analyst confirms it.
Illustrative bands, configurable per institution.
3 records linked. Requires analyst confirmation.
Illustrative demo data.
Interactive person, company, address, domain and document relationships with timeline filters.
Conflicts across sources shown side by side. Never auto-classified as fraud.
OCR and extraction for statements, filings, invoices and proofs, cross-verified against public data.
Claim, evidence, source, date, confidence. “Insufficient evidence” when support is thin.
No-code, versioned rules instead of a black-box score.
Alerts when directors, status, addresses or public records change.
Illustrative demo data. No real borrower or public record is shown.
Every finding keeps the original value, the model’s interpretation and the human judgement visibly separate.
Original value from a source with URL or reference, retrieved time, raw and normalised value, freshness and reliability class.
Model reasoning with confidence, evidence count and an explanation of why it was flagged. Always labelled as AI.
Human notes, resolution and status, attributed and time-stamped. Feedback tunes rules, never decisions.
| Field | Uploaded | Public record | Status |
|---|---|---|---|
| Company name | ABC Technologies | ABC Technologies | Match |
| Address | Mumbai | Pune | Review |
| Director | A. Mehta | Arjun Mehta | Probable |
| Incorporation year | 2019 | 2019 | Match |
Verify companies, directors, ownership and registry status with provenance.
Map borrowers, promoters and associated entities before the decision.
Surface duplicate identities, address concentration and document mismatches as signals.
Watchlists alert on director, status, address and public-record changes.
Compile cited evidence, timelines and approvals into a versioned report.
Extend the same graph and evidence to guarantors, vendors and beneficial owners.
Create investigations, search entities, review intelligence, upload documents, analyse borrowers and generate reports.
Investigate anomalies and connected entities, review adverse information, analyse patterns and record risk observations.
Review KYC and KYB, source provenance and screening results, inspect audit logs and configure compliance rules.
Run advanced OSINT, build entity graphs and timelines, save evidence and create investigation reports.
Review analyst cases, approve or reject internal findings, comment and request further investigation.
Manage users, permissions, data sources, API keys, retention, billing and templates.
Integrate investigations, signals, documents and reports into your loan-origination workflow. Connectors are pluggable; mock mode is clearly labelled.
# Reference API shape
POST /api/v1/investigations
POST /api/v1/entities/search
GET /api/v1/company/{id}
GET /api/v1/relationships/{id}
POST /api/v1/documents/analyze
POST /api/v1/ai/analyze
GET /api/v1/reports/{id}
GET /api/v1/signalsA reference architecture designed for governed data, explainable matching and auditable AI.
Every source implements one adapter contract carrying licence, rate-limit and reliability metadata. Requests pass a policy check, rate limiter, retrying fetch with circuit breaker, validation, normalisation and provenance-preserving persistence before indexing.
interface DataSourceAdapter {
id: string; type: SourceType;
terms: { licence; permittedUse[] };
limits: { rps; burst; dailyQuota };
reliability: 'official'|'verified'|'public'
|'news'|'user'|'ai'|'unknown';
fetch(q, ctx): Promise<RawRecord[]>;
normalise(r): CanonicalRecord;
}Normalise, block, compare, score, cluster, review. Features span name similarity, phonetic keys, address components, hard identifiers, graph context and time overlap. Analyst confirmations become must-link and cannot-link constraints.
score = Σ wᵢ · simᵢ(a, b)
confidence = calibrate(score)
band = Strong | Probable
| Possible | Weak
// never a confirmed identityEdges carry validity, confidence and evidence. Analytics include connected components, shortest path, community detection, centrality and motif detection, with a time slider and level-of-detail rendering for large networks.
MATCH (a:Company {id:$id})
<-[d1:DIRECTOR_OF]-(p:Person)
-[d2:DIRECTOR_OF]->(b:Company)
WHERE b <> a AND overlaps(d1.valid, d2.valid)
RETURN b, p, min(d1.confidence, d2.confidence)Query, intent and entity resolution, hybrid keyword and vector retrieval, graph expansion, tenant and role filtering, grounded reasoning, citation validation and guardrails. External content is treated as untrusted data. Uncited claims are blocked; thin evidence yields “Insufficient evidence”.
1 Query → intent + entities 2 Hybrid retrieval (keyword + vectors) 3 Graph expansion 4 Tenant / role access filter 5 LLM reasoning on retrieved items 6 Citation validator 7 Guardrails → answer
Policies are data. Conditions include thresholds, set overlap, graph metrics, time windows and source-conflict counts. Actions create signals or tasks, notify analysts or trigger webhooks. Every signal stores the rule version that created it.
{ "id": "dir-overlap-3", "version": 4,
"when": { "metric": "sharedDirectors",
"with": "otherBorrowers", "gt": 3 },
"then": { "signal": "Director overlap",
"severity": "review" } }Tenant ID on every record with row-level security, separate index and graph namespaces, per-tenant keys via envelope encryption, masked and tokenised PII with audited reveals, and an append-only, hash-chained audit log. SSO via SAML / OIDC, MFA, SCIM and RBAC with custom roles.
| Layer | Control |
|---|---|
| Identity | SSO, MFA, SCIM |
| Tenancy | RLS + per-tenant keys |
| Data | TLS, at-rest encryption |
| Audit | Tamper-evident log |
Users, cases, documents, reports and keys isolated per institution.
Masked by default, role-based reveals, every reveal audited.
Searches, exports, AI queries and permission changes recorded.
AI never approves, rejects, blacklists or declares fraud.
Source, retrieved time, raw and normalised values, freshness and reliability class. Facts, AI interpretation and analyst notes stay separate.
Multi-tenant SaaS, dedicated tenant, private cloud or VPC, with data-residency configuration and SSO via SAML / OIDC.
REST API and webhooks for investigations, signals, documents and reports. Pluggable data connectors and mock mode.
Flip through the executive edition or the full technical brochure without leaving this page.
| Capability | Typical point solution | Credmetric |
|---|---|---|
| Output | Lists of records | Connected entity workspace |
| AI answers | Free-text summaries | Cited, validated findings |
| Source conflicts | Ignored or silently resolved | Side by side, analyst-resolved |
| Scoring | Opaque vendor score | Transparent, configurable signals |
| Relationships | Static | Time-aware graph |
| Monitoring | Point-in-time check | Watchlists with change alerts |
| Sources | Fixed vendors | Pluggable, governed connectors |
| Decisions | Often implied | Human-in-the-loop by design |
Generalised comparison; individual products vary.
Authentication, organisations, universal search, company and person intelligence, entity resolution, graph, investigation workspace, document upload and OCR, AI summary and anomalies, reports, PDF export, audit log.
More connectors, watchlists, news and regulatory intelligence, bulk investigations, API and webhooks, analytics, advanced document analysis.
Enterprise SSO and SCIM, private and dedicated deployment, custom AI models, graph analytics, continuous monitoring, advanced institution rules.
Usage is based on investigations, entities, AI analysis, documents, API requests and reports. Pricing on request.
SSO, SAML, SCIM, dedicated or private deployment, data residency, custom retention and SLA.
Talk to usNo. It surfaces evidence, connections and flags for human review. Credit decisions stay with your authorised people and policies.
Permitted public sources, licensed providers and documents your institution uploads, through governed connectors with licence and rate-limit metadata. Sources are never presented as complete.
Answers are built from retrieved evidence and cite it. Uncited claims are blocked, and the assistant says “Insufficient evidence” when support is missing.
No. Demo and mock data are always labelled and never presented as real public records.
Yes. Institutions define versioned, no-code rules that create review signals, tasks or notifications.
Data minimisation, PII masking with audited reveals, purpose-of-processing and legal-basis fields, retention controls and delete workflows. It is designed to be configured to your institution’s applicable requirements; public data is not assumed to be unrestricted.
The architecture supports multi-tenant SaaS, dedicated tenants, private cloud or VPC deployment and data-residency configuration.
Other sources continue and the result states how many were reached, such as “3 of 5 sources queried”. Coverage is never overstated.
Tell us about your use case. The button opens your email app with a prefilled request.