Back to Portfolio

Shelfie

Smart inventory intelligence for households and small businesses. Computer vision meets inventory algorithms to track what you have, what you need, and when to buy.

Role Founder & Lead Engineer
Type B2B / B2C
Stack Python, Computer Vision, Mobile
Status In Development

See it, track it, predict it

Vision Layer

Camera-based shelf scanning identifies products, reads labels, and estimates quantities from a single photo.

Inventory Engine

Tracks consumption patterns, calculates depletion rates, and maintains real-time stock levels across locations.

Prediction System

ML-based reorder recommendations. Knows when you'll run out before you do and suggests the optimal buy time.

What it's built with

Python + FastAPI BACKEND
Computer Vision (YOLO) ML
React Native MOBILE
PostgreSQL + Redis DATA
Claude API AI
AWS S3 + CloudFront INFRA

Hard problems in inventory

Shelf recognition at scale

Training vision models to recognize thousands of product SKUs from phone camera photos in variable lighting conditions — not a lab environment.

Consumption prediction

Building accurate depletion models from sparse data. Households don't scan items out — the system has to infer usage from periodic shelf snapshots.

Mobile-first architecture

Heavy CV processing happens server-side, but the UX must feel instant. Optimistic updates and background sync keep the app responsive.

Multi-tenant inventory

Same system serves households tracking groceries and businesses managing stockrooms. Shared core, different rules and interfaces.