REAL-TIME INTERCEPT ACTIVE

Hardened Fraud Shield for Modern Rail.

Advanced ML-driven risk orchestration across UPI, IMPS, RTGS, SWIFT and AEPS. Sub-10ms anomaly scoring engineered for the velocity of India's payment networks.

CORE CAPABILITIES

Numbers that speak for themselves

AUISY currently manages structured and unstructured data, including billions of records.

0.0k TPS

UPI Load

₹0M

Prevented Loss / 24h

0.0%

False Positive Rate

01 — Multi-Protocol Defense

Unified visibility across every rail.

One control plane. Five protocols. Independent risk vectors per rail, correlated in real time across the entire transaction graph.

UPI

Unified Payments Interface

TPS

142.8k

LATENCY

4.2ms

FRAUD

0.018%

TREND

-2.1%

IMPS

Immediate Payment Service

TPS

28.4k

LATENCY

12ms

FRAUD

0.042%

TREND

-0.6%

RTGS

Real-Time Gross Settlement

TPS

1.2k

LATENCY

110ms

FRAUD

0.008%

TREND

stable

SWIFT

Cross-Border Messaging

TPS

842

LATENCY

2.4s

FRAUD

0.094%

TREND

+1.2%

AEPS

Aadhaar Enabled Payments

TPS

4.6k

LATENCY

340ms

FRAUD

0.061%

TREND

-0.3%

UPI · IMPS

Instant Domestic Rails

Real-time VPA behavior profiling, device fingerprint persistence, and collect-request risk filtering tuned for NPCI velocity.

• DEVICE FINGERPRINTING

• COLLECT-REQUEST FILTERING

• VPA REPUTATION

AEPS · MICRO-ATM

Last-Mile Biometric

Spoof detection on biometric capture and merchant-level risk scoring for Aadhaar-enabled rural cash-out networks.

• MERCHANT GEO-FENCING

• SPOOF PATTERN MODEL

• DUPLICATE-CAPTURE BLOCK

SWIFT · RTGS

High-Value Settlement

Enterprise AML screening and correspondent-banking risk for ISO 20022 cross-border and high-value domestic settlement.

• SANCTIONS SCREENING

• ISO 20022 ANALYTICS

• CORRESPONDENT-BANK SCORE

02 — Graph Engine

See the network, not the transaction.

Our neural engine surfaces hidden links between accounts, IPs, devices and VPAs. It doesn't just block transactions — it identifies the entire syndicate before the first packet is settled.

Link Analysis

Identify mule-account clusters and laundering paths across rails.

Identity Resolve

Correlate device IDs across AEPS, IMPS and UPI in a single graph.

Velocity Mapping

Detect coordinated fan-out attacks within a 200ms decision window.

03 — TEMPORAL ANOMALY MAP

72-hour signature heatmap.

Nominal

Elevated

Critical

REGION · IN-WEST · CORR-7

04 — Decision Pipeline

From wire to verdict in < 10ms.

01 · INGEST

Kafka stream from NPCI/SWIFT gateway. Schema-validated, signed at edge.

02 · ENRICH

Device, geo, VPA, historical-velocity vectors joined in-flight.

03 · SCORE

GNN + gradient-boosted ensemble emits a continuous risk score.

04 · DECIDE

Consensus engine returns ALLOW / REVIEW / BLOCK to the rail.

05 — How It Works

A six-step streaming flow, end-to-end.

AUISY is engineered around what a Chief Risk Officer or Head of Fraud actually answers for: identified, quantified, prevented, compliant, recoverable, defensible.

Edge ingress

NPCI, SWIFT and rail gateways stream into an at-the-edge Kafka topic with mTLS and signed envelopes.

Schema lift

ISO 20022, NPCI and AEPS payloads collapse into a single canonical transaction object.

Context join

Device, geo, VPA reputation, prior velocity and graph neighborhood are joined in <2ms.

Model ensemble

GNN + gradient-boosted trees + sequence model emit a continuous fused risk score.

Policy fusion

Risk score collides with bank-side policy: ALLOW, STEP-UP, REVIEW or BLOCK.

Online learn

Verdict outcomes loop back into the feature store inside the same minute window.

PREVENTIVE PLAYBOOKv24.11

→

Mule-account quarantine

Auto-freeze on second-degree graph match to a known mule cluster.

→

AEPS biometric replay defence

Reject identical minutiae template re-used inside a 24h window.

→

RTGS dual-control trigger

Mandatory second-officer approval for high-value beneficiaries.

→

Step-up auth on VPA drift

Force re-auth when payee VPA deviates from usual pattern.

→

Cross-border anomaly hold

Hold SWIFT transfers on anomaly or sanctions mismatch.

→

Velocity circuit-breaker

Cap UPI transaction bursts within rolling time windows.

07 — WHY AUISY

Built for the rails. Not retrofitted onto them.

Streaming risk fabric with graph + model native to rails.

CAPABILITY

TYPICAL

AUISY

Decision latency

200ms – 2s, often batch

< 10ms in-flight

Rail coverage

One or two rails, stitched

UPI · IMPS · RTGS · SWIFT · AEPS native

Detection logic

Static rules + thresholds

GNN + GBM + sequence ensemble

Identity model

Account-centric

Account · device · VPA · IP graph

Feedback loop

Weekly retrain, manual

Online, < 60s outcome → feature

Evidence trail

Free-text analyst notes

Cryptographically-signed decision log

Native graph core

Graph is runtime, not export.

Sub-10ms verdict

Built for ultra-low latency.

Five rails, one plane

Unified risk layer.

RBI-aligned

Mapped to controls 1:1.

Explainable scores

Transparent decisions.

VPC / On-prem

Secure deployments.

THE LEAKAGE BET

We will measure leakage on your live rail for 30 days. If AUISY doesn't beat your current stack, you don't pay.

Shadow-mode deployment. Same traffic, parallel verdicts, daily delta report signed by your audit team.

SHADOW WINDOW

30d

RISK TO BANK

₹0

IMPACT

<10ms

07 — OPERATING IMPACT

One platform. Three operating models.
Measurable P&L impact.

The same graph-native core adapts to banks, fintechs and payment infrastructure providers.

Banks

Protect deposits. Defend the licence.

↓ 62%

Fraud Loss

84%

Alerts Auto Closed

9.3x

Mule Discovery

< 6h

Dispute TAT

• Real-time scoring on UPI / IMPS / RTGS
• Auto-generated STR / CTR packets
• Branch-level risk dashboards
• Capital-risk modelling

Fintechs

Ship faster without inheriting fraud.

0.04%

Chargeback Ratio

↓ 41%

Drop-off

11 days

Go-live

₹38k

Revenue Saved

• Drop-in JS / mobile SDK
• Pre-tuned fraud packs
• Sandbox replay testing
• PCI-DSS tokenisation

Payment Infra

Keep the network honest at line rate.

240k TPS

Throughput

1,180+

Mule Rings Broken

↓ 91%

Leakage

8.4 ms

Risk Verdict

• Multi-tenant scoring
• Federated risk graph
• ISO-native integration
• Zero-replay deployment

CAPABILITY × OUTCOME MATRIX

How each capability lands on the balance sheet

MODEL v24.11 · 142 INSTITUTIONS BENCHMARK

Capability	Banks	Fintechs	Payment Infra
CAP-01 Graph-native entity resolution Resolves devices, VPAs, accounts and merchants into one identity across rails.	Detects synthetic identities before account opening — cuts onboarding fraud 71%.	One customer graph across PA, lending and wallet; KYC reuse drops cost 38%.	Cross-member mule scoring — surfaces rings invisible to any single bank.
CAP-02 Sub-10ms streaming inference GPU-accelerated XGBoost + GNN ensemble scoring on Kafka at line rate.	Blocks fraudulent UPI debit before NPCI deems-accepted — true prevention, not recovery.	No checkout latency penalty; conversion stays flat as fraud rules tighten.	Switch maintains SLA — risk adds < 1ms to round-trip at 200k TPS.
CAP-03 Explainable decisions Every verdict ships with SHAP attribution + rule lineage + graph evidence.	RBI Annexure-V audits close in days, not quarters; zero adverse findings YTD.	Customer-care can defend declines on call — complaint volume drops 54%.	Member-bank disputes resolved with shared evidence packet — no he-said-she-said.
CAP-04 Adaptive feedback loop Confirmed fraud and chargebacks re-train models nightly with drift guardrails.	Recovers from novel attack patterns in 24h vs 6-week vendor release cycles.	Rule packs auto-tune to your traffic — no in-house data-science team required.	Network-wide intelligence — one member's loss becomes everyone's defence.
CAP-05 Sovereign deployment On-prem, VPC or air-gapped; data never leaves the institution's perimeter.	RBI data-localisation compliant out of the box; no cross-border DPA exposure.	Pass bank-partner security reviews in one pass — accelerates sponsor onboarding.	National critical-infra classification supported; ISO 22301 BCP-aligned.