Mahalla

Mahalla — the first hyperlocal neighborhood services marketplace for Central Asia, powered by AWS DynamoDB, Vercel, and Gemma AI.

Comment

Inspiration

In Uzbekistan and Central Asia, there is no equivalent of TaskRabbit, Fiverr, or any hyperlocal services marketplace. If you need home-cooked plov from a neighbor, someone to do your laundry, or help moving furniture — you rely on word of mouth and manual searching through friends and family.

The concept of mahalla (махалля) — a traditional Uzbek neighborhood community where neighbors help each other — inspired us to build a digital version of this ancient tradition.

What it does

Mahalla is a hyperlocal neighborhood services marketplace where neighbors can offer and request everyday services:

  • 🍽️ Food — home-cooked meals, plov, samsa, bread
  • 👕 Laundry — washing and ironing clothes
  • 🛵 Delivery — neighborhood courier runs
  • 💪 Moving — heavy lifting and relocation help
  • 🧹 Cleaning — house cleaning services

AI Features:

  • Smart Match — users describe what they need in Russian or Uzbek, and Gemma AI instantly finds the best matching services nearby
  • AI Price Suggester — providers get fair market price recommendations in UZS based on the local Uzbekistan economy

How we built it

  • Frontend: Next.js 15 App Router, TypeScript, Tailwind CSS, shadcn/ui
  • Database: AWS DynamoDB (single-table design) — stores all services, orders, users, and reviews
  • Deployment: Vercel
  • AI: Gemma 4 31B via Google AI Studio API
  • UI Generation: v0 by Vercel

Challenges we faced

  1. DynamoDB schema mismatch — our initial table used serviceId as partition key but the code expected id. Solved by creating a new single-table design with id as the partition key.

  2. Gemma reasoning model parsing — Gemma 4 returns responses with a thinking block before the actual JSON. We had to extract the JSON from after the thinking block.

  3. VPN required — Vercel and v0.app are blocked in Uzbekistan, requiring VPN for all development work.

  4. AWS IAM permissions — initial deployment failed due to missing DynamoDB permissions for the IAM user.

What we learned

  • AWS DynamoDB single-table design patterns
  • Next.js 15 App Router with server components
  • Integrating Gemma AI for multilingual (Russian/Uzbek) natural language processing
  • Building for emerging markets with unique constraints

What's next

  • Real authentication with phone number (OTP via SMS)
  • Payment integration with Payme and Click (Uzbekistan payment systems)
  • Mobile app for iOS and Android
  • Expand to all cities in Uzbekistan and Central Asia
  • Real-time chat between customers and providers
Share this project:

Updates