● PMLA 2002 — STR Filing Deadline: 7 Days from Suspicion ● FIU-IND FINnet 2.0 — Active Integration ● RBI Data Residency — In-Region Processing Only ● MuleHunter.ai Alert Volume: ↑ 340% YoY ● DPDP Act 2023 — PII Handling Compliant ● AWS PrivateLink — Zero Public Internet Exposure ● PMLA 2002 — STR Filing Deadline: 7 Days from Suspicion ● FIU-IND FINnet 2.0 — Active Integration ● RBI Data Residency — In-Region Processing Only ● MuleHunter.ai Alert Volume: ↑ 340% YoY ● DPDP Act 2023 — PII Handling Compliant ● AWS PrivateLink — Zero Public Internet Exposure
Revolutionizing FinCrime Workflows

The Smart Workspace
for Financial Crime
Investigations

Collapsing bank compliance backlogs from 45 minutes to under 2 minutes with automated GSTN/MCA enrichment, graph-based mule-network detection, and FINnet 2.0‑ready agentic narrative generation.

🏦 Commercial Banks 🏦 Small Finance Banks 📱 Payments Banks 🔒 AWS PrivateLink Deployable 📋 FIU-IND FINnet 2.0
96%
Reduction in avg. validation time per alert
<90s
Async enrichment from GSTN + MCA + CBS
FINnet
STR narratives pre-formatted for FIU-IND 2.0
0 bytes
Customer data traversing public internet

Live Platform Preview

One Workspace. Every Context. Zero Tab-Switching.

The VigilAI analyst console surfaces entity enrichment, network topology, and a pre-drafted STR narrative in a single unified pane — from the moment an alert fires.

VigilAI Console — Investigation Mode — Analyst: priya.sharma@sbi.co.in
LIVE

Triggered Alerts

14 6 CRITICAL
● ACTIVE 09:42

Velocity Anomaly

A/C ···4821 — 38 txns / 4h

HIGH RISK UPI
● QUEUED 08:17

Sub-Threshold Structuring

A/C ···0093 — ₹49,800 × 9

MEDIUM NEFT
● QUEUED 07:55

Circular Fund Routing

5-node mule cluster detected

CRITICAL IMPS
● QUEUED 07:31

Dormant A/C Reactivation

A/C ···7734 — 18mo inactivity

MEDIUM CASA
● QUEUED 06:59

PEP Counterparty Match

A/C ···2219 — OFAC proximity 2

CRITICAL WIRE

Entity: RAJESH KUMAR GUPTA · A/C ···4821

5 linked nodes Circular path detected
A/C ···4821 R.K. GUPTA A/C ···0093 SHELL ENTITY A/C ···7201 A/C ···3348 A/C ···9912 A/C ···1105 ⚠ Circular route: 4 hops Subject Flagged node Neutral

✓ GSTN

Verified

GST filing gap: 3 qtrs

✓ MCA21

2 directorships

1 struck off entity

⟳ CBS

Loading 90d

38% complete

Agentic STR Draft

FINnet 2.0 Compatible
In-region LLM generating narrative…

STR Reference

STR/2024-25/VIG/004821

Reporting Entity

State Bank of India, Mid-Corporate Branch, Mumbai — IFSC: SBIN0001234

Investigation Summary

The account holder, Rajesh Kumar Gupta (A/C ···4821), exhibits velocity anomalies consistent with mule-account layering under PMLA Schedule offence categories. 38 transactions totalling ₹18,74,200 were processed over a 4-hour window on 14-Jun-2024, representing a 1,840% deviation from the 90-day peer-group baseline.

Graph-theoretic analysis reveals a 4-hop circular routing pattern across five accounts (···4821 → ···0093 → ···7201 → ···3348 → ···4821). MCA21 records confirm one co-director entity (DIN: 07312948) was struck off under Section 248 of Companies Act 2013. GSTN filings are in arrears for 3 quarters.

Recommended Action

File STR with FIU-IND — within 7 days (PMLA §12)

Simulated analyst console. Entity names and account numbers are illustrative. In production, all data remains within your bank's AWS VPC.

The Market Gap

India's AML Stack Has a Broken Handoff

Detection engines fire alerts at machine speed. What happens next is entirely human, entirely manual, and operationally broken.

Speed 1

The Trigger Engine

MuleHunter.ai & TMS Rules Fire in Milliseconds

RBIH's AI detection engine and internal transaction monitoring system (TMS) identify anomalous patterns — velocity spikes, sub-threshold structuring, PEP proximity — and generate alerts in near real-time.

Automated Alert Queue Population

Alerts are scored, prioritised, and routed to the Financial Crime Investigation Unit (FCIU) queue within seconds. The machine's job ends here.

High and Rising Alert Volumes

With UPI transaction volumes exceeding 14 billion/month, alert queues at commercial banks often contain 200–800 open items at any given time, with Payments Banks experiencing the sharpest growth.

System latency from transaction to alert:

< 2s end-to-end detection

Speed 2 — Broken

The Validation Bottleneck

Manual GST Portal Lookup (7–12 min)

Analyst opens GSTN portal in a separate tab, searches by PAN/GSTIN, manually screenshots and copy-pastes filing history and turnover data into a local case notes file.

Manual MCA21 Director Lookup (5–8 min)

Separate login to MCA21 v3 to pull DIN records, company charge sheets, and cross-directorships. No API — strictly UI-driven. Results are copy-pasted into investigation notes.

CBS Ledger Pull Request (10–15 min queue)

Analyst submits a report request to the CBS team for 90-day ledger history. This often sits in a shared mailbox queue with SLAs measured in hours, not seconds.

STR Narrative Typed from Scratch (15–20 min)

With all data finally assembled across 6+ tabs, the analyst manually drafts a PMLA-compliant investigation narrative and reformats it for FINnet 2.0 XML upload. High error rate, low consistency.

Avg. analyst time per alert validation:

45 min of fragmented manual effort
VigilAI collapses this gap to < 2 minutes

Core Capabilities

Three Engines. One Investigation Workspace.

Each capability is architected to address a specific layer of the investigation workflow — enrichment, analysis, and filing — so your analysts never leave the platform.

Module 01

Asynchronous Context Enrichment

The moment an alert is opened, VigilAI dispatches parallel API calls to GSTN, MCA21 v3, and your Core Banking System simultaneously. Results converge into a single enriched entity dossier in under 90 seconds — no context switching, no waiting, no copy-pasting.

GSTN Filing Status Live API
MCA21 v3 Director Graph Live API
CBS 90-Day Ledger Core Banking Connector
KYC / CKYC Repository Configurable
Module 02

Relational Linkage Graphing

Money-mule networks are graph problems. VigilAI builds a directed transaction graph per entity — surfacing circular fund routing, fan-out layering, and shared-identifier clusters (mobile, device ID, nominee DIN) that are invisible in tabular ledger views.

Circular path detection Graph traversal
Fan-out / fan-in analysis Centrality scoring
Community clustering Louvain algorithm
Interactive node expansion Analyst-driven
Module 03

Agentic STR Narrative Generation

A sovereign, in-region LLM synthesises enrichment outputs and graph findings into a structured investigation narrative mapped to PMLA Schedule offences and FATF typologies. Output is pre-formatted for direct upload to FIU-IND's FINnet 2.0 portal — analyst reviews, edits, and submits.

PMLA Schedule mapping Automated
FINnet 2.0 XML schema Pre-formatted
In-region LLM inference No data egress
Analyst edit + approval Human-in-loop

Workflow

From Alert to Filed STR in Under 2 Minutes

01

Alert Fires

MuleHunter.ai or TMS generates a risk alert. VigilAI receives the event via secure webhook within your VPC.

02

Context Assembled

Parallel async jobs pull GSTN status, MCA21 directorships, and 90-day CBS ledger. Enriched dossier ready in <90s.

03

Graph Rendered

Transaction network visualised. Circular routes, mule clusters, and shared identifiers flagged automatically for analyst review.

04

STR Filed

Agentic LLM pre-drafts the PMLA narrative. Analyst approves and submits directly to FIU-IND FINnet 2.0 portal.

Core Architecture & Security

Designed for RBI Data Residency Requirements. Not as an Afterthought.

VigilAI deploys entirely within your bank's existing AWS environment — every compute, inference, and graph workload runs inside a private network boundary in AWS Asia Pacific (Mumbai) ap-south-1. Customer financial data never traverses a public network, never touches a shared cloud tenant, and never leaves Indian soil.

Amazon Bedrock & Bedrock Agents

Sovereign GenAI Execution

All LLM inference and agentic STR-drafting pipelines run exclusively via Amazon Bedrock in the ap-south-1 (Mumbai) Region. Bedrock Agents orchestrate the enrichment-to-narrative workflow with full tool-use calling — no external model endpoint is ever contacted. Maintains 100% compliance with RBI's domestic data localisation mandate.

Inference region ap-south-1 only
Agentic orchestration Bedrock Agents
External model calls None — zero egress
RBI data localisation ✓ Satisfied
Amazon Neptune

High-Performance Linkage Analytics

Real-time node-and-edge relationship maps — tracing multi-layered shell companies, circular fund routes, and shared-identity clusters — are powered by a fully managed Amazon Neptune graph database instance. Neptune's SPARQL and openCypher engines enable sub-second traversal across millions of entity relationships without ETL overhead.

Database engine Amazon Neptune
Query languages openCypher · SPARQL
Traversal latency < 200ms typical
Deployment mode Fully managed / HA
Amazon VPC & AWS PrivateLink

Perimeter Isolation

The entire Copilot orchestration layer — every enrichment call to CBS, GSTN, and MCA21, every graph query to Neptune, every inference request to Bedrock — communicates exclusively through AWS PrivateLink endpoints inside the bank's Amazon VPC. No request ever traverses the public internet. No VPN tunnel required.

Network boundary Amazon VPC
Service connectivity AWS PrivateLink
Public internet traversal None — 0 bytes
VPN dependency Not required

AWS Service Map

ap-south-1 · Mumbai

Amazon VPC — Bank Private Subnet

Amazon Bedrock

Agents · Model inference

ap-south-1 ✓

VigilAI App Layer

ECS Fargate · Private subnet

Isolated compute ✓

Amazon Neptune

Graph DB · Multi-AZ HA

openCypher · SPARQL ✓

Amazon RDS + S3

Case store · AES-256 · KMS

Encrypted at rest ✓

AWS PrivateLink Endpoints

Core Banking

↕ VPC Endpoint

GSTN API

↕ VPC Endpoint

MCA21 API

↕ VPC Endpoint

AWS CloudTrail

Immutable audit log · 5yr retention

AWS KMS

Customer-managed keys · CMK

Every service-to-service call flows through AWS PrivateLink. Zero public internet traversal. Traffic stays on the AWS private backbone end-to-end.

RBI Data Residency — Fully Satisfied

All inference, graph traversal, and storage workloads execute within AWS ap-south-1 (Mumbai). Satisfies the RBI's 2018 Payment System Data Storage circular and subsequent localisation guidance, with Amazon Bedrock enforcing in-region model execution at the infrastructure layer.

DPDP Act 2023 — PII Isolation by Design

Customer PII is processed inside dedicated ECS Fargate task environments scoped per investigation, with AWS KMS customer-managed keys for field-level encryption. No cross-tenant data leakage. Audit logs retained 5 years via CloudTrail + S3 Object Lock — PMLA Section 12 compliant.

No Third-Party Model Calls — Ever

STR narrative generation and enrichment reasoning are handled exclusively by Amazon Bedrock foundation models with cross-region inference disabled. Investigation data never contacts OpenAI, Anthropic's API, Google Vertex, or any off-AWS endpoint — enforced at the VPC network policy layer.

Immutable Audit Trail via AWS CloudTrail + S3

Every analyst action, Bedrock inference call, Neptune graph query, and enrichment event is written to AWS CloudTrail with S3 Object Lock (WORM) enabled. Tamper-proof investigation records for FIU-IND examinations and internal audit — 5-year retention by default.

AWS

AWS Activate Portfolio — Deep Native Integration

Credit-Eligible Architecture

VigilAI's infrastructure is built entirely on AWS-native managed services — no third-party databases, no off-cloud model endpoints, no hybrid licensing complexity. Every workload is metered directly through the AWS billing console, making credit application and burn-down reporting straightforward for the review board.

Amazon Bedrock Bedrock Agents Amazon Neptune Amazon VPC AWS PrivateLink ECS Fargate Amazon RDS AWS KMS AWS CloudTrail Amazon S3
Limited Private Pilot — Q3 2025

Your analysts deserve better than 14 browser tabs.

We're onboarding a select cohort of Indian banks for our private pilot programme. If you're a CISO, CCO, or Head of Financial Crime Operations at a commercial, small finance, or payments bank — let's talk.

Or email us directly at solutions@vigilai.ai