Skip to content
All Systems
SYS_02 / 15SHIPPED • Enterprise ML Stack
FLAGSHIPFeaturedPublic • TypeScriptEnterprise Full-Stack & Applied ML

FraudShield

Real-Time Enterprise Banking Fraud Detection Platform

End-to-end real-time enterprise banking fraud detection platform with a 99.93%-accuracy ML model, live transaction streaming, 3-tier RBAC (Analyst/Investigator/Admin), device fingerprinting, anomaly tracking, and case management.

Next.js 16React 19TypeScriptJava 21Spring Boot 4PythonFastAPIKerasSupabase Realtime/RLSTailwind CSS v4

Model Accuracy

M-01

99.93%

Keras ML Classifier

Access Control

M-02

3-Tier RBAC

Analyst/Inv/Admin

Core Services

M-03

3 Runtimes

Next.js/Spring/FastAPI

Telemetry

M-04

Live Stream

Supabase Realtime/RLS

Section 01

Problem Statement

Banking fraud operations require low-friction coordination between live transaction ingestion, ML risk scoring, device fingerprinting, and role-scoped case investigation.

Section 02

System Constraints

  • >Separate high-throughput transaction rules/RBAC from Python ML scoring
  • >Enforce strict 3-tier row-level access control across Analyst, Investigator, and Admin roles
  • >Stream live anomalous transactions and case updates to investigators

Section 03

System Architecture & Dataflow

Topology // 4 Nodes • 3 Edges
01

Transaction & Device Stream

Ingest
02

Java 21 Spring Boot 4

Score
03

FastAPI + Keras (99.93%)

Realtime Alert
04

Next.js 16 + Supabase RLS

Section 04

Key Engineering Decisions