SinkAlert Rail · Verity Technology · รฟม.

End-to-End Data Flow

How รฟม. feeds data → how sources are pulled → how the Risk Engine works → how every user gets the result. All hosted in our AWS cloud (SaaS).

flowchart LR
    subgraph SRC["1 · DATA SOURCES"]
        direction TB
        MRTA["รฟม. (MRTA)
alignment · chainage · asset
borehole · piezometer · settlement
TBM log · construction · incident"]
        OPEN["Open sources
Sentinel-1 InSAR (Copernicus)
DMR geology · DGR groundwater
TMD rain · BMA GIS"]
    end

    subgraph CLOUD["2 · SINKALERT SaaS — AWS (Verity-hosted)"]
        direction TB
        ING["Ingestion
API Gateway · file upload · data room"]
        PULL["Scheduled ETL pull
(auto-refresh)"]
        CAT["Data Catalog
lineage · quality · PDPA"]
        RISK["Risk Engine
MCDA + XGBoost/LightGBM
InSAR · SHAP"]
        DB["Data store
PostgreSQL · object storage"]
    end

    subgraph OUT["3 · DELIVERY"]
        direction TB
        DASH["Web Dashboard
Executive · Engineering · Field"]
        API["Read-only API
OpenAPI · OAuth2"]
        ALERT["Alerts
LINE OA · email · in-app"]
        RPT["Reports
PDF · Excel · CSV"]
    end

    MRTA -- "upload / API" --> ING
    OPEN -- "auto-pull" --> PULL
    ING --> CAT
    PULL --> CAT
    CAT --> RISK --> DB
    DB --> DASH
    DB --> API
    DB --> RPT
    RISK -- "risk events" --> ALERT

1 · How data enters

รฟม. uploads / pushes via API / secure data room (alignment, borehole, TBM, incident…)
Open sources are auto-pulled on schedule (InSAR, geology, water, rain, BMA)

2 · Inside the SaaS (AWS)

Ingestion → Catalog (lineage + quality + PDPA) → Risk Engine → store
We host all data; รฟม. never runs its own infra

3 · How users get data

Dashboard (web) · read-only API · alerts (LINE/email/in-app) · reports (PDF/Excel/CSV)
Field inspectors get mobile-responsive task + photo upload