GABA Water Intelligence

From water signals to verified environmental action.

Comment

Inspiration

Water loss is often discovered after damage has already occurred. Leaks, abnormal pressure, reservoir drawdown and failing infrastructure can waste scarce water long before the problem becomes visible.

GABA Water Intelligence is designed to help operators see those signals earlier, understand the evidence behind them, and move from environmental data to accountable action.

What it does

GABA is a water-intelligence MVP for environmental and infrastructure monitoring.

The prototype combines:

  • simulated flow, pressure, reservoir and pump-health telemetry
  • a live water-system command centre
  • operational map context
  • digital-twin views
  • deterministic anomaly detection
  • evidence-backed AI risk briefs
  • explicit human Approve / Amend / Reject decisions
  • inspection-ticket creation after human approval
  • evidence and audit trails
  • environmental and operational reporting
  • a guided leak-to-impact demo

The core principle is:

AI detects, predicts and recommends. Humans approve consequential actions.

How the demo works

The demo begins from a normal operating baseline in a fictional South African water district.

GABA then injects a deterministic leak pattern:

  1. Night flow rises beyond its expected band.
  2. Downstream pressure falls.
  3. Reservoir drawdown accelerates.
  4. GABA correlates the signals and creates a high-confidence leak alert.
  5. An evidence-backed risk brief explains the sensor sources, timestamps, reasoning and projected impact.
  6. A human operator reviews the recommendation.
  7. The operator approves a priority field inspection.
  8. An inspection ticket is created for the field team.
  9. A simulated repair is verified.
  10. The network returns to baseline and the system records the outcome.

The final 2.34 ML/day figure shown in the demo is a projection based on the simulated detected loss rate of 2,340 m3/day. It is not a measured real-world municipal result.

Architecture

Sensors / APIs / GIS / weather -> ingestion -> environmental data layer -> time-series analysis -> digital twin -> anomaly detection -> GABA intelligence -> risk scoring -> human decision gate -> inspection / action -> evidence and audit.

The current hackathon build uses a deterministic simulator so the full workflow can be demonstrated without physical field hardware. The adapter boundary is designed for future LoRaWAN, MQTT, Modbus and HTTP telemetry sources.

Human-governed infrastructure AI

GABA does not autonomously isolate valves or perform consequential infrastructure control.

The system provides decision support. A human reviews the evidence and authorises the operational response.

This keeps the workflow explainable, accountable and auditable.

Earth Forward impact

GABA is focused on water conservation and climate resilience.

The long-term goal is to help water operators:

  • detect water loss earlier
  • prioritise high-risk interventions
  • understand why an anomaly was detected
  • connect alerts to accountable field action
  • maintain traceable evidence of environmental interventions
  • measure the operational outcomes of those interventions

What existed before NextStep Hacks

Before this hackathon, the project existed as an earlier AEI Water Intelligence MVP.

The repository history shows that the earlier build already included:

  • simulated water telemetry
  • a map-based command interface
  • a digital-twin view
  • evidence-backed AI risk briefs
  • a human decision workflow
  • inspection tickets
  • evidence and audit views
  • reporting
  • a production-oriented schema
  • architecture and pilot KPI documentation
  • CI validation

What I built during NextStep Hacks

During NextStep Hacks 2026, I focused and extended that earlier MVP into GABA Water Intelligence.

The hackathon work included:

  • rebranding and repositioning the water-intelligence product as GABA
  • creating the new environmental editorial landing experience
  • building a guided judge-friendly normal-to-leak-to-impact demo
  • starting the demo from a clean normal baseline
  • deterministic leak injection
  • dynamic anomaly detection and critical alert creation
  • evidence-brief progression inside the guided demo
  • human approval linked to inspection creation
  • simulated field repair verification
  • post-intervention return-to-baseline logic
  • a projected avoided-loss impact view
  • hackathon demo documentation
  • Cloudflare-ready static build tooling
  • public deployment of the working prototype
  • final brand and interface cleanup

This disclosure is included to clearly separate the pre-existing foundation from the work completed for this hackathon.

Built for demonstration, designed for deployment

The current application uses fictional demo sites and simulated telemetry. It does not claim a live municipal deployment or measured water-saving result.

The purpose of this MVP is to demonstrate the complete operating model:

water signal -> anomaly -> evidence -> human decision -> inspection -> verified action -> environmental outcome.

Built With

Share this project:

Updates

Submission history