← All work
IoT + AI

Smart Self-Checkout Basket

An IoT smart basket with sensor fusion and a companion app that lets shoppers skip the checkout line entirely — hardware through to app.

Client
Retail-tech startup
Discipline
IoT + AI
Engagement
Full remote product team
Sensor fusion
barcode/RFID + weight + ultrasonic + camera in one device
Hardware to app
full stack delivered by one remote product team
Skip the line
checkout completes as items go into the basket

Context

A retail-tech startup set out to remove the checkout line in busy airport-retail environments, where time-pressed travellers abandon purchases rather than queue.

The challenge

Delivering a true self-checkout experience meant solving it end to end — physical sensing hardware, reliable item recognition, and a frictionless payment app — as one integrated product, on startup timelines.

Our approach

Sense the cart

We integrated barcode/RFID, weight, ultrasonic sensors, and camera modules into a smart basket, fusing their signals to know what's in the cart in real time.

Recognize and track

Item recognition and cart tracking keep an accurate running tally as items go in and out.

Pay and walk out

A companion mobile app handles in-app payment and self-checkout, so the shopper never joins a queue.

End to end, as one team

A full remote product team delivered hardware, firmware, recognition, and app as a single integrated build.

DEVICE / FIELDEDGECLOUDBasket SensorsRFID/weight/cameraOnboard Sensor FusionItem-event detectionCompanion AppCart + paymentCloud BackendSession + inventory sync
Sensor fusion runs on the basket itself for low latency; the cloud backend reconciles cart state, payment, and inventory

Architecture

Sensor fusion inside the basket, not a single sensor doing everything

No single sensor type reliably solves item identification at checkout — barcode scanning alone misses items placed without a clean barcode-facing orientation, weight alone can't distinguish two items of similar mass, and camera-only systems struggle with occlusion when a basket fills up. The basket combines barcode/RFID reading, weight sensing per item placement, ultrasonic proximity sensing to detect item add/remove events, and camera modules for visual confirmation — with onboard compute fusing these signals to determine, with high confidence, what was added or removed and when.

Onboard processing with a thin sync layer to the app

Running all sensor fusion in the cloud would introduce latency that breaks the 'just walk out' experience — by the time a cloud round-trip confirms an item was added, the shopper has already moved on. Sensor fusion and item-event detection run on the basket's onboard compute, with the resulting cart state synced to the companion mobile app in near real time. The app is where the shopper sees their running total and completes payment; the basket's job is purely to maintain an accurate, low-latency cart state.

A backend that reconciles cart state with payment and inventory

The cloud backend's role is reconciliation: linking a basket's session to a shopper's account, maintaining the authoritative cart record, handling payment on checkout, and — for the retailer — feeding accurate basket-level data back into inventory systems. Edge cases (an item added then removed, a sensor-fusion low-confidence event needing app-side confirmation, a basket going out of connectivity range mid-session) are handled with a reconciliation protocol between basket and backend rather than assuming the edge state is always correct.

What we built

  • A multi-sensor smart basket (barcode/RFID, weight, ultrasonic, camera)
  • Onboard sensor-fusion firmware for real-time item-event detection
  • A companion mobile app for cart view and payment
  • A cloud backend for session management, payment, and inventory sync
  • An edge-to-cloud reconciliation protocol for connectivity and confidence edge cases

Technology stack

Hardware / IoT
Barcode/RFID readersWeight sensorsUltrasonic proximity sensorsCamera modulesEmbedded firmware
AI / Edge
Sensor fusionOn-device item-event detectionComputer vision (item confirmation)
Mobile
Companion app (iOS/Android)Real-time cart sync
Backend
Cloud session & payment serviceInventory integrationEdge-cloud reconciliation

Results & impact

The result was a working self-checkout experience that removed the line entirely — proof that a small senior team can deliver an integrated hardware-plus-software product from the ground up.

  • The retail-tech startup got a working hardware-to-app product from a single remote team, avoiding the integration risk of separately sourcing hardware, firmware, app, and backend from different vendors.
  • Sensor fusion meant the basket could maintain high-confidence cart state without requiring shoppers to scan items individually — closer to a true 'just walk out' experience than barcode-only alternatives.
  • On-device processing kept the shopper-facing experience responsive regardless of connectivity quality, with the cloud backend handling reconciliation rather than being a real-time dependency.
  • The architecture was built to support pilot deployments first, with the edge-cloud reconciliation protocol designed so the system degrades gracefully rather than failing hard when connectivity or sensor confidence is imperfect — important for a product still being validated in real stores.

Have a similar problem to solve?

Tell us what you're building. We'll tell you the fastest honest path to shipping it.

Start a conversation →