LIFELINE: Mind the Makan Gap

Inspiration

Go to any neighbourhood hawker centre at 7am and you'll see the same faces: uncles and aunties with their kopi, at the same place as yesterday. For many of them it's not just breakfast. It's the one affordable meal within walking distance, and often the main conversation of the day.

Then a notice goes up: closed for renovation, back in three months. Most of us just walk somewhere else. But what about the auntie who lives alone upstairs and can't walk far?

We realised nobody actually measures this. Singapore counts hawker centres per 10,000 residents, which treats every centre and every resident the same. So we wanted to build a tool that shows, for every hawker centre, how many seniors and lower-income residents have no other centre within walking distance, and what happens to them when it closes.

What it does

LIFELINE ranks every hawker centre by how many seniors and lower-income residents depend on it, meaning they have no other centre within walking distance. Planners can then "close" any centre on a map to see who is left without an affordable meal nearby, and which centres absorb them. It helps NEA stagger renovations, and helps town councils and grassroots groups plan interim support.

How we built it

This is an idea-stage submission, so this is our planned build. Open data from data.gov.sg and SingStat is ingested with Lakeflow into Delta tables governed in Unity Catalog. The dependence score and closure simulation run in SQL and notebooks, with score versions tracked in MLflow. Outputs are an AI/BI dashboard, a Genie space for plain-English questions, and a Databricks App for the simulator.

Challenges we ran into

Since this is still pre-liminary building stage, we didn't encounter a lot of challenges yet other than those related to datasets. The closure data has no permanent closures, so we are considering for our trend view to only show long renovations. The open waste dataset covers total waste, not food waste, so that view needs a second source. The data covers hawker centres only, not coffee shops/kopitiam, and we use straight-line distance. We state these limitations openly.

Accomplishments that we're proud of

We confirmed every dataset the core idea needs before pitching it. The key find was SingStat's subzone table by age and dwelling type, which lets us measure dependence far more finely than planning area. We also found a real case in the data: three neighbouring centres in Ang Mo Kio renovated at different times.

What we learned

A sharper question beats more generic features. Asking "who cannot do without this centre?" gave us one metric that ties together access gaps, closures and vulnerability. We also learned to check the data before committing to using that dataset.

What's next for LIFELINE: Mind the Makan Gap

If shortlisted, we plan to build the pipeline, score, gap map and simulator in the two-week sprint. After that: an add-a-centre mode to test sites for new hawker centres, AI-written briefs for each area, and adding coffee shops/kopitiams if open data allows.

Built With

• data.gov.sg
• databricks
• databricks-apps
• databricks-sql
• delta-lake
• genie
• geojson
• lakeflow
• mlflow
• python
• sql
• unity-catalog