العربية
All work

SE201 · 2026 · Leader and backend developer

Rent & Build

A construction equipment rental marketplace for Saudi Arabia, built so two renters can never book the same machine for the same dates.

Designed as a team in SE201 at Prince Sultan University (6 students). I led the team and built the backend, with Claude Code as an AI pair programmer.

  • Spring Boot
  • React
  • MySQL
  • k6
Rent & Build equipment page: an excavator illustration next to the booking card with rental dates, the total price and a Request booking button
Before
2,743 ms
After indexes
259 ms

Search p95 · 100,000 listings · k6, 20 virtual users

FIG. 01 · Rent & Build · Search p95
Rent & Build three-tier architecture: owner and renter interfaces, then messaging, authentication, booking and search services, then user, vehicle and booking databases.

Architecture by the SE201 team

FIG. 01b · Rent & Build · Architecture
Rent & Build home page with equipment search by type and city, and recently listed machines
FIG. 01c · Rent & Build · Home page

Screen recording of Rent & Build: browsing equipment, requesting a booking, the owner approving it, and the renter paying.

FIG. 01d · Rent & Build · Demo

Problem

Equipment owners have no shared place to list machines, and renters call around. The costly failure is a double booking.

What I built

Owners list machines and approve requests. Renters search, compare, book, pay and message owners. Admins manage users and see an audit log.

Key decisions

  • Bookings lock the vehicle row inside a transaction. A test fires 10 simultaneous approvals for overlapping dates on real MySQL, and exactly one wins, on every CI run.
  • I measured search before changing it. EXPLAIN-driven indexes and a capped result count took search p95 from 2.7 s to 259 ms on 100,000 listings (k6, 20 virtual users, one 4-core VM).
  • Payments are verified on the server, and a database constraint allows only one successful payment per booking. Payment flow verified against a mocked Moyasar API.
  • On a simulated free hosting tier the app took about 4.5 minutes to start, so I chose an always-on host for the demo.

Results

  • Search p95: 2,743 ms to 259 ms
  • Throughput: 16.6 to 164.5 requests per second at 2x load
  • 91 unit and 96 integration tests on real MySQL, plus an end-to-end test
  • 94% line coverage on services