Quiet Doors

Inspiration

As a Silver Generation Ambassador, I've done house visits to seniors in Geylang Serai. Outreach like this is organised by address: volunteers work through a list, block by block, with no way of knowing which doors matter most. I remember visiting a senior living with her children and grandchildren but you can tell from her eyes that she wasn in risk of social isolation.

Singapore is now super-aged: 21.4% of citizens were 65 or older in June 2026. The number of seniors living alone more than tripled between 2010 and 2025, to 88,400. But living alone isn't the whole story. A Singapore study of 16,943 older adults found that only 14.4% of socially disconnected seniors lived alone; 78.8% lived with family. Lists of seniors living alone miss most of the people outreach is meant to find. I wanted to help volunteers decide where to look first.

What it does

Quiet Doors ranks Singapore's planning areas, subzones and HDB blocks by social isolation risk, so Silver Generation Ambassadors and Active Ageing Centre staff can visit the highest-risk blocks first.

Coordinators start from a map of planning areas, drill down to subzones, and see the 10 highest-risk HDB blocks in each, with the signals behind each ranking. They can also ask questions in plain English through Genie, such as "Which blocks in Geylang are highest-risk?"

Quiet Doors doesn't claim to identify isolated individuals; no open data can do that. It narrows where human outreach should look first. Seniors in private estates are covered at subzone level; HDB areas are ranked down to individual blocks

How we built it

(Planned for the 2-week sprint)

• Data: four open datasets:
  • SingStat's June 2026 residents by subzone, age, sex and dwelling type
  • Census 2020 household size by planning area
  • Census 2020 household income from work by planning area
  • HDB Property Information, which gives rental flat counts for about 13,300 blocks
• Pipeline: Databricks notebooks load each dataset into Delta tables in Unity Catalog. HDB blocks are geocoded and joined to their subzone and planning area.
• Risk index: a transparent index combining age, household, income and housing signals, weighted using published Singapore research.
• Robustness checks: the index is cross-checked against a PCA-derived ranking and tested with ±20% weight shifts. Every run is tracked in MLflow.
• Output: an AI/BI dashboard with a drill-down map and ranked block lists, plus a Genie space for plain-English questions

Built With

• databricks