AIDA Intelligence Hub · AIDA Consulting

How this working model works

This is the AIDA Intelligence Hub baseline running on synthetic data. On this page every number is calculated in your browser, when you click, by the same KPI logic that the live working model runs as SQL in PostgreSQL. The sections below describe that live model. The records themselves are fictional.

Construction started: September 26, 2026

The path of one click

Your browserAsks for a KPI, a period and a product
Web serverChecks your sign-in, adds a secret service token, forwards the request
Secure gatewayAdmits only requests that carry the token; everyone else is refused
Data platformA read-only API runs the governed KPI function in PostgreSQL and returns the result

1. The data platform

Plays the part of the client's data platform. Everything runs in containers as its own project, with its own network, storage and passwords. It shares nothing with any other system.

PostgreSQL 16: the reporting mart
PostgREST: the read-only web API
A secure tunnel connector: the only way in
A database browser: reachable only on the platform itself

2. The PostgreSQL database

Built in three layers, the way a production mart would be:

mart: 41 tables of fictional records (customers, alerts, cases, regulatory filings, screening hits, reviews, training, feed loads, data-quality checks)
kpi: 50 governed SQL functions, one per KPI. The same function feeds the tile, the chart, the report and the CSV, so the numbers can't disagree.
api: a handful of read-only functions. These are the only thing the web can reach.

3. The read-only API

The web role can call the API functions and nothing else. It cannot read a single table directly.
The functions run as a reader account that can see the mart and change nothing.
Every query runs in a read-only transaction with a 15-second limit.

4. The secure gateway

The data platform makes an outbound-only, encrypted connection to Cloudflare. No inbound port is opened, and the database is never directly reachable from the internet.

The API sits behind Cloudflare Access, which admits only requests carrying a service token held by the web server.
It is separate from every other AIDA system.

5. The web server

A separate cloud server hosts the page with nginx.

Its API route adds the service token on the server side, so the token never reaches a browser.
It forwards read requests only.
It stores no data. It holds the web page and passes requests through.

6. A year of fictional data

12 months: September 25, 2025 to September 24, 2026, plus 60 days of earlier history so queues look normal on day one
About 10% of real volume for a bank this size: 1,500 customers across Rewards credit card, Online savings, TFSA and GIC
Generated from a fixed seed, so the same data can be rebuilt exactly at any time
No client data: every customer ID, case and figure is invented

7. Checked end to end

Before going live, every dashboard value and every report (summary and records) was compared between PostgreSQL and an independent reference engine across 44 period and product combinations.

2,244 payloads compared, 0 differences

8. If the data platform is offline

The dashboard carries a built-in offline copy of the same model. If the API doesn't answer, it switches over automatically, and the chip next to "As of" changes from Live PostgreSQL via API to Offline copy.

How it maps to a client's build

In this working model	At the client
A containerized data platform	The client's data platform (Azure, Databricks, Synapse or Fabric)
PostgreSQL `mart` schema	AML Reporting Mart: read-only, reporting only
50 functions in the `kpi` schema	Certified KPI functions, one per KPI, under version control
Cloudflare Tunnel + Access service token	Power BI data gateway or private network link
A separate web server	Power BI Service, with row-level security by role
Fictional records	Daily feeds from monitoring, cards, deposits, KYC, screening and FINTRAC