PersonalPersonal projectFull-stack engineer2024 — 2025
QuickEMart
AI quick-commerce: describe a meal, get a cart.
A quick-commerce platform with 7,000+ products, 170+ brands and 40+ categories, where a natural-language prompt generates the cart. MongoDB compound indexing, Razorpay payments and Maps-based delivery.
7,000+
products
170+
brands
01
Why I built it
Grocery apps make you search item by item. I wanted to type “breakfast for two” and get a sensible cart.
02
The problem
- Fast filtering across a large catalogue.
- Turning a prompt into real, in-stock products.
03
System design
- 1
React client
Prompt-to-cart flow and catalogue browsing.
- 2
Node.js API
AI cart generation, payments and delivery.
- 3
MongoDB
Compound indexes for low-latency filtering.
04
Decisions & trade-offs
Compound indexes shaped around the filters
Why: Category, brand and price filters hit an index instead of scanning 7,000+ products.
Trade-off: Indexes cost write time and memory; only the real filter paths get one.
05
Backend & frontend
Backend
- Razorpay payments and webhooks; Google Maps delivery workflows.
Frontend
- Natural-language cart generation UI.
06
Results
- Working AI cart generation across the full catalogue.