Inspiration

For generations, hawker centres have been a cornerstone of Singaporean life, not just as places to enjoy affordable, multicultural food, but as vibrant social spaces where communities connect. It's no surprise they hold a special place in the hearts of so many Singaporeans. This living cultural fabric was officially recognized on December 16, 2020, when Singapore's hawker culture was inscribed on the UNESCO Representative List of the Intangible Cultural Heritage of Humanity, solidifying its place in the nation's identityYet this living heritage faces growing pressures.

Data from data.gov.sg underscores a critical decline: active NEA hawker stalls fell from 13,430 in 2023 to 12,943 in 2024, and then plummeted to 11,446 in 2025—representing an 11.6% drop in just 12 months. Coupled with temporary closures and unequal spatial distribution, this sharp contraction severely jeopardizes elderly and lower-income residents who critically depend on these spaces for daily sustenance, financial relief, and community belonging which drove us to choose KopilamAI

Our mission and vision

Mission: To empower NEA with unified data intelligence to proactively identify, prioritize, and safeguard vulnerable hawker centres across Singapore.

Vision: A resilient urban landscape where every community maintains uninterrupted access to affordable food, social connection, and living heritage.

Challenges we ran into

A primary challenge in this initial phase was determining how to unify fragmented open-data sources (spanning NEA stall counts, closure schedules, and SingStat demographic datasets) into explainable, actionable metrics. Beyond the technical data integration, we had to navigate the complex stakeholder landscape across civic agencies to ensure KopilamAI’s vulnerability index aligns with real-world decision-making frameworks and translates directly into targeted policy interventions.

Accomplishments that we're proud of

We are proud of transforming fragmented public data into a focused, actionable decision support solution. Rather than forcing every available dataset into the model, we evaluated data relevance, removed weak proxies and concentrated on indicators that could genuinely help NEA prioritise hawker centres and target interventions. We also designed the solution’s Databricks architecture from scratch despite having no prior experience with the platform.

What we learned

We learned how to structure an end-to-end Databricks solution and transform geospatial and demographic data into decision relevant indicators that could support NEA. We also learned to prioritise data relevance over data quantity. Some datasets such as recycling bin locations, did not meaningfully represent hawker food waste, while available food waste data from Hawker centres lacked the granularity needed for actionable NEA metrics.

Intended architecture

Lakeflow will ingest NEA and SingStat open datasets into a Bronze, Silver, Gold Delta Lake architecture governed by Unity Catalog. PySpark / SQL notebooks clean and integrate the data and calculate accessibility, closure impact and vulnerability indicators into a hawker-centre priority index. Gold tables power the Databricks App and AI/BI dashboards, while Genie enables non-technical stakeholders the ability to explore data through natural language queries