Inspiration

I own a small bodaboda fleet in Dar es Salaam. I am also a second-year emergency medicine resident at Muhimbili National Hospital.

The rent-to-own arrangements I ran with my own drivers lived in a paper exercise book. Every dispute came down to the same thing: the owner's memory against the driver's memory, with no shared record between them. A driver who had paid faithfully for eighteen months owned nothing he could show anyone — no history, no proof, no asset.

Across Tanzania that pattern repeats hundreds of thousands of times. The missing thing was never credit. It was a ledger both sides could see.

What it does

LUNO records marejesho — rent-to-own repayments between fleet owners and the drivers buying motorcycles and three-wheelers from them.

Both parties see the same balance on their own phone. The driver builds a verifiable payment history he owns. The owner stops chasing.

LUNO lends nothing and charges drivers nothing: daftari, si mkopeshaji — the ledger, not the lender. That positioning is deliberate. Every framework imported into this market assumes the operator wants to become a lender. The locally material question was different: who holds the record, and does the person paying trust it?

Kiongozi — the AI layer

Four Gemini agents on Cloud Run run the operations a solo founder cannot:

  • Marejesho Reminder — repayment nudges in Kiswahili, timed to each driver's earning pattern rather than to a calendar date
  • Risk Flagger — surfaces slipping accounts before they become defaults
  • Owner Weekly Digest — a plain-language fleet summary written for each owner
  • Application Scorer — assesses new driver applications against that owner's own criteria

These run unattended, on schedule, in production. They are not a demo layer bolted on for this competition — they are how the business runs while I am on a clinical shift.

How I built it

Solo, while working full-time in an emergency department.

  • Android — Kotlin / Jetpack Compose (com.lunomj.driver, live on Google Play)
  • iOS — Swift / SwiftUI (live on the App Store, id6782275114)
  • Backend — Supabase / PostgreSQL
  • Agents — Python on Google Cloud Run, calling Gemini via Vertex AI
  • SMS — Beem Africa
  • Automation — GitHub Actions

Challenges

Payment infrastructure. Tanzanian cards are refused by most cloud providers. Getting billing to work at all required routing through Selcom and Payoneer.

Constrained devices. Drivers use low-end Android phones on expensive data. Everything degrades gracefully or it does not ship.

No CS background. I am a physician who learned to build by building. Frontier AI is my technical co-founder. That is not a caveat — it is the thesis. If domain experts can now ship production software, the software gets built by people who actually understand the problem.

Accomplishments

  • Registered as LUNO Mwanajando Technologies Co. Limited (BRELA 206903325)
  • PDPC-certified for data protection in Tanzania
  • Two live app store listings
  • Active across 11 Tanzanian regions
  • Named, consenting customers available for verification

What's next

Insurance and fuel distribution over the same rail. We own the ledger — distribution is the product.

Built With

Share this project:

Updates