CDU IT Code Fair 2026 · AI Challenge · Scenario 1

Fair Fix

Housing repairs for remote Northern Territory communities, ranked by risk to the household, not by how far away the house is.

Tenants report with pictures and a few taps. A transparent scoring tool ranks every job, an AI double-checks the report, and a person always makes the final call.

Try the live demo Report as a tenant
  • 5 days official urgent target for a remote home, vs 2 in town
  • 39.8% of Indigenous households in remote NT are overcrowded
  • 0 public records of how long repairs actually take

What we are solving

Remote tenants wait longer, and nobody can see it

The NT looks after 5,498 remote homes in 73 communities and town camps. Repairs are booked by category, remote targets are written to be twice as long as town ones, and the department says it does not track how long repairs really take. That leaves a chain of delays:

  1. Hard to reportPhone forms in English. Prepaid phones, patchy coverage, and over 100 local languages.
  2. Ranked by categoryA sparking power point and a sticky door can land in the same queue.
  3. Wait for a tripRemote jobs wait for the next contractor visit. Targets are already 2× longer.
  4. No feedbackActual completion times aren't recorded, so the system can't learn or improve.

This matters legally too: in the Santa Teresa case (Young v CEO (Housing) [2023] HCA 31), the High Court held that tenants can be compensated when their landlord fails to keep a remote home habitable.

What we found

The data behind the problem

Figure 1 Official repair targets: remote homes are given twice as long Remote Town
Urgent
5 days
2 days
Routine
28 days
14 days

+14 daysA routine repair in a remote home is allowed two extra weeks, by policy, before it even counts as late.

Immediate (life-safety) jobs are 4 hours everywhere. Source: NT Estimates Committee, QON 7.10, June 2025.

Figure 2 Overcrowding rises sharply with remoteness Indigenous households in Indigenous-managed social housing that are overcrowded, 2022
8.2%
Major city
8.3%
Inner regional
19.4%
Outer regional
32.7%
Remote
49.3%
Very remote

More people per house wears out plumbing, hot water and laundries faster. Source: AIHW, Housing assistance in Australia 2023, Table SUITABILITY.2.

Figure 3 Remote NT compared with Greater Darwin Aboriginal and Torres Strait Islander households and people, 2021 Census Rest of NT Greater Darwin
Household is overcrowded
39.8%
10.5%
Rents public or community housing
64.6%
21.8%
Speaks an Indigenous language at home (people)
75.4%
9.2%

Remote tenants depend on the public landlord, live in fuller houses, and mostly speak a language other than English. Source: ABS 2021 Census, Aboriginal and Torres Strait Islander Peoples Profiles (tables I05, I13, I16).

Show the figures as a table
MeasureRemoteTown / citySource
Urgent repair target5 days2 daysNT QON 7.10
Routine repair target28 days14 daysNT QON 7.10
Overcrowded (SOMIH, very remote vs major city)49.3%8.2%AIHW 2023
Overcrowded (Rest of NT vs Greater Darwin)39.8%10.5%ABS 2021
Renting public or community housing64.6%21.8%ABS 2021
Indigenous language at home75.4%9.2%ABS 2021
$43.1mspent on remote maintenance in 2022, about $7,843 per house
51.5%of that spend coded only as "misc / other", so nobody can see what it fixed
23.9%of remote houses got a condition inspection in 20 months
100+Aboriginal languages and dialects spoken across the NT

Spend and inspections: Grealy, Su & Thomas (2025), IJERPH 22(6):836. Languages: NT Aboriginal Interpreter Service.

How we address it

One finding, one fix

Hard to report in English text

Picture-led reporting

Tenants tap a picture of the problem, answer yes / no / not sure, and add photos. It works on a cheap phone, and a housing officer can lodge it for them.

Jobs ranked by category, not risk

Safety-first priority index

Life-safety hazards (power, gas, sewage, no water) always go first and are never pushed back for distance. Then household need and time against the official target.

Remote jobs lose to travel cost

Fairness you can see

Trips to one community are shared across all its jobs. The coordinator sets how much distance counts and sees, job by job, what that choice changes.

Reports disappear with no answer

A contact at every stage

Every ticket has a timeline, the official due date and a named person to call, from lodged to fixed and confirmed by the tenant.

Nobody measures real repair times

Promised vs actual

Contractors promise a time for each area. The system records when the job was really done and uses that history to suggest who to send next.

AI and sensitive data

Trusted AI by design

Our own tool computes the score. The AI only sees a redacted report: no names, places, contacts or health details. It can raise a job's priority, never lower it, and a person decides.

Built against the NT Government AI Assurance Framework

  • Community benefit Faster fixes for the households most at risk.
  • Safety Life-safety rules are fixed, not learned.
  • Fairness The equity trade-off is shown and logged.
  • Privacy Personal and health data never leave our system.
  • Transparency Every rank comes with plain-English reasons.
  • Accountability Every AI suggestion and human decision is recorded.

Our data

Where our numbers come from

There is no public dataset of NT repair requests, so the demo uses synthetic jobs built on the official categories and targets. Every figure on this page comes from a public source:

Who we are

The team

IT students at Charles Darwin University, building for the CDU IT Code Fair 2026 AI Challenge.

  • Bigyan Siwakoti
  • Binay Siwakoti
  • Shashwat Acharya
  • Abishek Kandel

Try the live demo

All people, homes and jobs in the demo are made up. Please don't enter real personal details.

Coordinatormike / pass1234Triage queue, dispatch, needs-attention dashboard
Tenantrosie / pass1234Report a problem with pictures and track it
Contractortop / pass1234Accept jobs and update progress
Log in to the demo Create a tenant account