Skip to content
All case studies

Internal tool · Computer Vision

Techneth Attendance

An in-office attendance terminal for Techneth. Employees tap check in or out and look at the camera; the server runs liveness checks and face-descriptor matching, then records the punch in Firestore. A fingerprint-sensor bridge, leave and holiday workflows, an approved out-of-office timer, and HR reports complete the admin side.

Role
Sole developer
Client
Techneth
Timeline
Aug 2026
Year
2026
T

Face recognition with blink liveness, matched only on the server

Fingerprint identification through a local sensor bridge

Leave, holidays, reports, exports, and HR email in one admin

Overview

Techneth needed attendance that was quick for staff and hard to game. The earlier approach, detecting phones on the office Wi-Fi, could be fooled and said nothing about who was actually present, so I replaced it with a face- and fingerprint-based kiosk.

What I built

A Next.js app that runs on an office machine: a kiosk screen for check-in and check-out, and an admin for enrolment, leave, calendars, reports, and exports, backed by Firestore. Face recognition runs on the server with face-api on TensorFlow.js; the kiosk only sends frames, and the server performs the liveness check and the match. A small Python service bridges a USB fingerprint sensor over loopback with a bearer token.

Outcome

Staff check in by looking at a screen, admins get leave, holidays, and reports in one place, and every punch is tied to a verified face or fingerprint.

Key features

Kiosk captures a short burst of frames; the server checks face geometry and blink liveness before matching 128-d descriptors

Identity is decided server-side, so a tampered kiosk page can't punch in for someone else

Admin face enrolment and fingerprint 1:N enrol and identify via a token-protected Python bridge

Out-of-office work timer that requires admin approval

Leave requests, holiday calendar, bulk remarks, reports, CSV export, and HR mail

Runs on an office Ubuntu machine under systemd, discoverable on the LAN

Built with

Next.jsFirebase / Firestoreface-api (TensorFlow.js)PythonNode.jssystemd

Have a project like this?

I take products from first sketch to shipped, design, build, and the AI in between. Tell me what you’re working on.

More work