SilverWatch

Inspiration

Active Ageing Centre outreach is still largely manual, so coordinators cannot see which planning area to visit first. In 2026, 21.4% of Singapore citizens are aged 65 and over, up from 20.7% in 2025. That crosses the UN line for a super-aged society. National totals do not say where to go.

What it does

SilverWatch scores each planning area with an Outreach Priority Index. It averages three published inputs with equal weight: the share of residents aged 65 and over, the share of one-person households (all ages, a proxy, not elderly-only), and eldercare coverage, where less coverage ranks higher. Coordinators get a heatmap and a top-10 list for the week. It does not identify people, and it is not a trained model.

How we built it

This round is the concept. If shortlisted, the build is a two-week sprint on Databricks Free Edition: ingest public CSV and GeoJSON in a notebook, store Delta tables in Unity Catalog, join them in SQL, log the index in MLflow, and publish an AI/BI dashboard. The score is a SQL expression, so a coordinator can see why an area ranked where it did.

Challenges we ran into

The latest planning-area census is from 2020. The 2025 and 2026 figures are national context only. There is no current Active Ageing Centre address file, so coverage uses Senior Activity Centre points from MOH Eldercare Services.

Accomplishments that we're proud of

The index uses only published, area-level data, names its limits up front, and does not score individuals. The weights stay equal until coordinators can check them.

What we learned

A useful outreach list does not need a black-box model. An auditable average, with the gaps written down, is enough to change which area gets visited this week.

What's next for SilverWatch

If shortlisted, build the notebook, the logged index, and the dashboard. Not in this sprint: a 2030 forecast, a recommender for new centres, or an app.

Built With

• databricks
• delta-lake
• deltalake
• mlflow
• sql
• unity-catalog
• unitycatalog