Skip to main content
Back to selected work

Civic Health / Offline Mobile App

HEALTH

A civic health app that helps residents decide whether to seek emergency help, find a nearby facility, or continue to a lower-acuity next step—even with unreliable connectivity.

HEALTH safety check warning residents to contact emergency services when someone is in immediate danger.
Source: Public repositoryThe symptom flow begins with a deterministic safety check so immediate-danger guidance appears before any assisted triage.

Public evidence

These reviewed public sources support the adjacent build claims; they do not imply adoption or unmeasured outcomes.

Problem

Target user
Naga City residents deciding whether symptoms call for emergency help, a nearby facility, or a lower-acuity next step.
Challenge
A symptom tool needs to react predictably to immediate danger while still making supported assessments, medications, and facility information useful when connectivity drops.
Why it mattered
An urgent case should never wait for a generative answer, and losing locally needed information during an outage would undermine the app when guidance matters most.

Constraints

  • The product was built by a five-person team within a hackathon timeline.
  • Offline support covers defined assessment, medication, and facility records—not every care scenario.
  • Emergency and mental-health inputs must receive deterministic handling before any eligible request reaches Gemini.

Solution

The app handles immediate safety first, uses Gemini only for eligible lower-risk inputs, and keeps a defined set of care information on the device.

  1. Residents answer an immediate-danger and mental-health check before any AI request can run.
  2. Inputs that clear that check can receive Gemini-assisted triage and guidance toward local facilities.
  3. Supported assessments, medication records, and facility data stay in SQLite and use synchronization when connectivity returns.

Engineering Decisions

  1. Route danger signals before calling Gemini

    Constraint
    Emergency and mental-health inputs cannot depend on model availability or model-generated wording.
    Decision
    Run deterministic emergency and mental-health checks before eligible requests reach Gemini.
    Rationale
    Give urgent inputs a predictable response path before assisted triage is considered.
    Trade-off
    Only inputs that clear the deterministic checks are eligible for Gemini-assisted triage.

  2. Keep a defined record set on the device

    Constraint
    Residents may need assessment, medication, and facility information without reliable connectivity.
    Decision
    Persist those records in SQLite and synchronize them with the backend when connectivity returns.
    Rationale
    Keep the supported care information readable during an outage without overstating universal offline coverage.

Outcomes

Product and delivery

  1. Top 15 / 200+

    Hackathon result

    The five-person team reached the semi-finals, placing in the top 15 of 200+ teams at the 1st Naga City Mayoral Hackathon.

  2. Offline records

    Care data kept on-device

    Assessments, medication records, and facility data remain available locally, with synchronization when connectivity returns.

Learnings and Next Steps

What I learned

  • In a health product, AI is a conditional helper—not the first or final authority for urgent safety.
  • “Works offline” is only credible when the exact records available without a connection are named.

What I would improve

  • Bring clinicians and residents into structured review and field testing before treating the guidance as production care infrastructure.

What remains unvalidated

  • The hackathon placement demonstrates team delivery, not clinical effectiveness, resident adoption, or production reliability.