العربية
All work

Alenmaa Hackathon · 2026 · Backend developer and tester

Mubsir

An AI tool that finds the cheapest Saudi online store for a product, and an honest look at where it failed.

Team project, Alenmaa Hackathon 2026. My part: backend, testing and statistical evaluation.

  • Spring Boot
  • SerpAPI
  • LLM
  • Statistical evaluation

Coverage

84.8%

28 of 33 test items found

Matching precision

78.1%

95% CI: 63.8% to 92.4%

Average price difference

2.5%

on correct matches

Accessory picks

50%

Share of cheapest-store picks that were accessories, not the product

FIG. 03 · Mubsir · Evaluation

What we built

A Spring Boot backend that searches stores through SerpAPI and uses an LLM to match listings to the product.

Results

  • It found the product for 28 of 33 test items (84.8% coverage).
  • 78.1% matching precision (plus or minus 14.3 points, 95% confidence interval).
  • 2.5% average price difference on correct matches.

What I found

Half of the “cheapest store” picks were accessories, like cases and cables, not the product itself. I proposed a relative statistical filter to remove them.