DillyDally SilverWatch

Team Name: DillyDally

Members:

Dylan Long Kai Xi @ Tian Cheng | National University of Singapore | Year 2 Business Analytics | [email protected]

Lee Zhi Yan, Natalie | National University of Singapore | Year 3 Business Administration and Computer Science | [email protected]

Inspiration

Singapore's elderly population is passing 1 million, yet outreach to isolated older adults is still largely manual and its coverage uneven. Talking to the problem statement, we noticed that the real bottleneck isn't a lack of maps. It's a prioritisation decision: with a fixed number of social workers and volunteer hours, where should outreach go next, and how do we justify it?

A population map answers "where do many elderly live?" It cannot answer "where does likely need most exceed the support within reach?" Active Ageing Centres (AACs) each serve a defined area, so a block can be close to one centre yet belong to another's catchment, or to none. We built SilverWatch around that gap.

What it does

SilverWatch ranks planning areas by an Outreach Priority Index and turns the ranking into a weekly outreach round for each social worker, scoped to their own AAC's catchment. Every stop comes with a plain-language reason. Coordinators get a heatmap and a top-10 list with a stability score. It is an area-level decision-support signal, never a statement about any individual.

How we built it

Data (all open): HDB Elderly & Future-Elderly Population, SingStat Resident Population 65+ by Living Arrangements and Key Indicators on the Elderly, and MSF's Active Ageing Centres and ComCare households, plus planning-area boundaries.

Databricks pipeline: files land in a Unity Catalog Volume, then flow through Delta Bronze → Silver → Gold. PySpark handles cleaning, area-key mapping, geocoding and block-to-AAC assignment. MLflow logs every scoring run. An AI/BI dashboard serves coordinators, a Databricks App serves workers, Unity Catalog limits each worker to their own catchment, and Lakeflow Jobs reruns the pipeline end to end.

Challenges we ran into

• To be updated.

What we learned

• To be updated.

What's next

Centre-siting recommendations, a 2030 demographic pipeline view, and validation with practising social workers and volunteer coordinators.

Built With

• databricks
• powerpoint