Simat Labs

DCI — Deep Customer Insight: On-Device Facial Recognition for Cafes and Retail

AI SDKv.1.13

DCI — Deep Customer Insight

A state of the art, real-time facial recognition solution for retail, education, and airports.

Talk to Sales

Khaled Alaa

{ name: "Khaled Alaa",
  last_order: "Cafe Latte with Soy Milk",
  number_of_visits: 12 }

DCI (deep customer insight) is a facial recognition solution, which recognizes repeat or new customers and their order history.
It gives retailers the ability to identify retail customers and give them highly personalized experiences.

DCI utilizes a network of edge devices across multiple locations and cloud infrastructure to recognize customers in realtime.The on edge machine learning systems use facial recognition along spatial intelligence to identify customers, in less than 200 milliseconds.

On-device

Inference runs locally on the POS or a small edge box. Faces never leave the store.

<200ms at POS

Returning guest detected before the cashier finishes a greeting. No network round-trips.

Privacy-first

No biometric uploads. Built around opt-in, on-device embeddings — GDPR-friendly by design.

Offline first.

DCI runs entirely on-device — no cloud round-trips, no biometric uploads. Works the moment your POS boots, even when the network doesn't.

Sub-200ms at the POS

Recognize returning guests in under 200ms — fast enough to greet them before they reach the counter.

Privacy by Design

All recognition happens on-device. No biometric data ever leaves the terminal.

4.5%

Returning Guest Insights

Track repeat visits, dwell time, and loyalty patterns without tying data to identity.

Drop-in with your POS

Connects to your existing POS stack — no new hardware, no workflow changes for staff.

Dashboard preview

Multi-Location Ready

Run independently at every store or sync recognition across locations — your call.

Unmatched accuracy.

Depth-aware sensing locks onto the guest standing at the counter and ignores everyone else.

Live targeting
  • Guest at the register
  • Diners at tables
  • Staff behind the line
  • Passersby on the street

Available across the region

DCI in Saudi Arabia, the UAE, and Jordan

Frequently asked

How does DCI identify returning customers?

DCI generates a privacy-preserving face template the first time a guest is enrolled, then matches new captures from a till-mounted camera against the local template store on an on-premise compute unit in under 200 milliseconds. Raw images are not sent to the cloud, and templates never leave your environment unless you explicitly enable cross-location sync. Depth-aware sensing locks onto the guest at the register and ignores staff, diners at tables, and passersby on the street.

Is DCI GDPR compliant?

DCI is designed to support GDPR — face templates are generated and matched on-premise, raw images are not uploaded, and biometric data does not leave your environment by default. The system supports purpose limitation, data minimization, the right to erasure, and lawful-basis enrollment, with signage and staff-training templates aligned to common European requirements. Actual compliance depends on how you configure DCI in production (retention, signage, lawful basis, DPA), so we work alongside your legal team during pilot rollout.

Which POS systems does DCI integrate with?

DCI integrates with common POS and loyalty systems — returning-guest signals can surface in the cashier interface, push to a loyalty platform to attach order history, or be exposed via a local API for kitchen display systems and CRM tooling. Recognition runs on-premise, so the integration is a local network call rather than a cloud webhook, with no internet round-trip at the till. If your POS is not yet supported, Simat Labs scopes the integration during pilot kickoff and typically validates a single-location rollout end-to-end within one to two weeks of hardware arrival.

What hardware does DCI need at the POS?

DCI runs on lightweight on-premise hardware — a standard USB or IP camera at the till plus a small fanless mini PC or industrial edge box running Linux, which handles detection, embedding, matching, and the local POS API on-device. No GPU is required for typical cafe and retail throughput, though one can be added for very high-traffic stores. There is no separate cloud server to provision and no biometric data is uploaded, so deployment is mostly mounting the camera and pointing the compute unit at your local network.

Last updated: