Context
Attendance in Iranian classrooms meant either a teacher reading names aloud for several minutes of every session, or a hardware scanner each site had to buy, mount, and maintain. Timez removed both: the teacher takes one photo of the room, and image processing marks every student present.
The reframe
Everyone in the category was competing to sell schools a better scanner. But a scanner is a thing a school has to buy, mount, power, and repair, and the camera was already in the room. It was in the teacher's pocket, already out, already being used to photograph the whiteboard. Attendance was never a hardware problem. It had just been sold as one for long enough that nobody was checking. Once you see that, the product designs itself: don't sell them a device, use the one they already carry.
What I built
A photo-based attendance product for teachers, deployed across universities, SAMPAD schools, private schools, and every branch of Kanoon Ghalamchi, Iran's largest supplementary education network. Most of the real design work was not in taking the photo; it was in the correction path, because the model is wrong often enough that a teacher's trust depends on how cheap it is to fix a mistake, not on how rarely one happens.
What I killed
What I killed was the central image store. Face matching worked well, well enough that the temptation was to keep every photo and mine it later. But these are children's faces. So instead of one central store, the system ran on each school's own local servers and auto-deleted the images within short windows. That cost us the longitudinal data a cleverer product could have used, and I made that trade on purpose: for minors, the safest data is the data you never keep.
Outcome
The proof was adoption, not a decimal I can quote. Roll call stopped eating the first minutes of every class. Sites no longer had to buy or maintain a scanner. And it spread the way a tool spreads when it genuinely helps: from single classrooms to whole institutions, including every branch of the country's largest tutoring network. When a thing keeps getting used long after the novelty wears off, that is the outcome.
Where my role stopped
I was product director, designer, and I managed the team, but the vision model itself was not mine. Engineers built and tuned the image processing; my job was deciding what the product should do when the model was uncertain or wrong, and how a teacher would come to trust it. I take credit for the product judgment, not the computer vision.
What I know now
The most ambitious version of Timez is the one that never shipped. It was going to plug into the Ministry of Education's Nemad wellbeing program and, over time, read the emotional state of a class with AI, quietly flagging the students who were struggling and needed someone to notice. It did not die because the idea was wrong; it died in the administrative layers above us. What I learned is that in education the ceiling on a product is often bureaucratic, not technical, and the best thing you design is not always the thing you are allowed to ship.