A web system that records when institute staff arrive and leave using their fingerprints - and, for teaching staff, how long each lecture actually ran.

The problem

An institute measures two very different things with the same word, "attendance". Office staff are measured on hours at work. Faculty are measured on lectures actually taken. A single in-and-out register cannot do both. On top of that, attendance feeds payroll, so a quiet correction by a clerk can quietly change somebody's salary.

What we built

• Two attendance models in one system - non-teaching staff punch at the gate or office reader, and the system works out arrival, departure and real working hours. Teaching staff punch inside the classroom: one punch opens a lecture, the next closes it, and the system records the room, the subject and the actual duration against the timetable.
• Maker-checker approval - a clerk can enter and correct anything, but nothing they type touches the live record. The change is held as a proposal, an administrator sees a clear before-and-after comparison, and only approval makes it real.
• Staff register - identity, designation, department and bank details, with role-based access so not everyone sees everything.
• Holiday and leave calendar - so Sundays, declared holidays and approved leave do not show up as mass absenteeism.
• Timetable - the planned lectures each lecture punch is measured against.
• Eight kinds of report - daily, monthly, late-comers, lecture-wise and more, each downloadable as Excel or as a printable PDF for signing.
• Full audit trail - who changed which record, from what to what, and when.

How it helps

• The salary register can be prepared from real worked hours rather than a manual muster.
• Lecture delivery is measured against the timetable instead of being self-reported.
• Corrections are still possible - they simply cannot happen without an approval on record.
• Every report can be printed and signed, which is what an audit actually asks for.

Technology

A web application over a Node and Express API with a relational database of twenty-four tables. Passwords are hashed with argon2, sessions use short-lived tokens that rotate, and the attendance engine is covered by an automated test suite. The system is designed to read from fingerprint readers, and ships with a simulator that drives the same engine so the whole flow can be tested before hardware is installed.