System Architecture

The underlying cloud infrastructure powering the BankGuard platform.

This platform converges two independent backend pipelines into a single, unified read layer.

By sharing a common schema design, the presentation layer (this Next.js application) is completely decoupled from the heuristics that generate the data.

1. Compliance Auditor

An AWS Lambda function triggered on a chron-schedule by EventBridge. It continuously audits the active AWS environment against a curated subset of the CIS AWS Foundations Benchmark.

When misconfigurations are detected (e.g., exposed S3 buckets, unencrypted EBS volumes), findings are formatted and persisted to the shared DynamoDB table.

2. Fraud Monitor

A high-throughput batch pipeline processing transactional data drops in an S3 bucket.

Each transaction is evaluated by a dual-layer engine:

Heuristic Rules: Fast evaluation of hard constraints (velocity limits, merchant categorization).
Machine Learning: An Isolation Forest model trained to detect multi-dimensional anomalies in the PCA space.

Anomalous transactions are persisted to the shared DynamoDB table, with critical incidents immediately dispatched via SNS.

3. The Shared Data Layer

Both pipelines converge on a single DynamoDB table utilizing a Single-Table Design.

This Next.js dashboard requests data through a lightweight API Gateway endpoint (the "Lambda Lith" pattern), ensuring the frontend remains thin, performant, and solely focused on presentation.