SLIDE 1

Problem & why it matters

HeatGuard · DAISI Challenge Singapore 2026 · Round 1

Team: KoolAid Problem statement: B1 – Urban heat risk mapping

Heat and haze warnings tell us it is hot. They do not tell us who is most at risk, or where cooling help is missing.

1 in 5

residents will be aged 65 or older as Singapore becomes officially “super-aged” in 2026 - NPDT

1M+

elderly residents, the group most at risk from heat stress.

2026

Transboundary haze remains a recurring annual threat in Southeast Asia, with Singapore at high risk of severe haze in 2026. - CNA

Source: DAISI Challenge Singapore Participant Guide

Why it matters now: no tool joins live heat, elderly density and distance to a cooling refuge to show where to act first. HeatGuard does, so outreach goes where it helps most.

SLIDE 2

Solution & data

HeatGuard · DAISI Challenge Singapore 2026 · Round 1

HeatGuard: a live Heat Risk Score that ranks planning areas by where it is hot, elderly residents live, and no cooling refuge is nearby.

WHO USES IT, AND THE DECISION IT CHANGES

Volunteer coordinators and community workers at Active Ageing Centres. It tells them where to send outreach first and which centres to open as cooling refuges on a hot day.

DATASETS (ALL ON DATA.GOV.SG)

1. Air temperature, real-time API – NEA

2. Rainfall, real-time API – NEA

3. HDB elderly & future-elderly population – HDB

4. Active Ageing Centre locations – MSF

OUR DIFFERENCE

The refuge-gap risk score

Generic dashboards show where it is hot. HeatGuard shows where it is hot, many elderly residents live, and the nearest Active Ageing Centre is far away.

Heat

live, NEA

Elderly

density, HDB

Refuge gap

nearest AAC

→ Low / Medium / High per planning area, a ranked top 10 and a “why this score” panel

SLIDE 3

Databricks architecture & impact

HeatGuard · DAISI Challenge Singapore 2026 · Round 1

1 Sources

NEA live APIs, HDB, MSF

2 Ingest

Lakeflow → Delta Lake

3 Transform

Clean, join by planning area

4 Model

Calculate Risk for each area

5 Output

AI/BI dashboard + Genie

Two live streams (temperature, rainfall) plus two static datasets. Unity Catalog lineage and access controls as a stretch goal.

OUTPUT WE WILL DEMO

A live dashboard: heat map, Low / Medium / High risk by planning area, a top-10 priority list, a “why this score” panel and adjustable weights.

MEASURABLE IMPACT IF IT WORKED

Outreach triage moves from guesswork to a ranked top 10. We measure it by backtesting past hot days and beating an elderly-density-only baseline. No health-outcome claims.

REALISTIC IN TWO WEEKS

Core: live ingest, risk score, dashboard and MLflow tracking on Free Edition. Stretch: next-day forecast, Unity Catalog. The score is a composite index, as no heat-harm labels exist.