That was the tagline on our pitch deck in 2018. Two computer science undergrads at a liberal arts college in Lahore, a Raspberry Pi in a plywood box we cut ourselves, and a claim that sounded absurd coming from us: that we could build a doorbell which recognised your face and let you in.

It won Best of Punjab in Embedded Systems at NUST’s COMPPEC’18. It never became a company. Both of those facts taught me something, and it took me the better part of a decade to work out what.

What it actually was

S.E.R.I.C. — Smart Embedded Redundant IoT Camera.

A Linux-based single board computer running the entire software suite locally, on the board itself. OpenBR embedded on the system side handling facial recognition. The same board managing remote connections on request. Face-based entry, remote unlock for guests, prerecorded messages for when nobody was home.

The framing we opened every pitch with was this: people have been opening doors the same way since doors were made. Someone knocks. Someone else has to be home. That person has to get up, walk over, and look. It is slow, it requires presence, and it has not meaningfully changed in centuries.

We checked what already existed — Ring, Chui, Amazon Key, Abode. Some of them streamed video. One of them unlocked doors. None of them did facial recognition. That gap was the entire thesis.

The enclosure was plywood. The demo ran off a ThinkPad on a classroom desk. It looked exactly like what it was: two students who had figured out something real with almost no budget.

The part I was actually good at

I was not the one writing most of the code.

What I did was own the product. The idea, the positioning, the pitch, and the refusal to let it stay a class project. For a long time I treated that as the lesser contribution — the thing you admit to quietly. I no longer think that is right.

We finished the build ahead of schedule. Instead of coasting into the deadline, I pushed it into rooms where it could be tested as a business rather than as an assignment: the Kazan OIC Entrepreneurship Boot-camp, the Telenor Velocity IoT Hackathon. It placed in the top hundred internationally. Our final report’s own conclusion says it plainly — we came back from those rooms convinced the thing had market potential, and we were moving in that direction.

You do not win a category at an engineering school on silicon alone. Somebody has to stand in front of a judge who has seen forty projects that day and make them care about the forty-first. That was my job, and I was good at it.

Why it stopped

I kept pushing after graduation. I genuinely believed in it.

Then I got seriously ill, and that is where it ended. Not funding. Not a falling-out. My body set the timeline, and there was no argument to be made with it.

I have made peace with that, mostly. What I have not made peace with is how long I let it stand as a reason to stop building anything at all.

The bit that gets me now

On the last slide of that 2018 deck, under a heading that just said Future, we wrote:

tailor made experience through ML

I had no idea how to do that. None. It was a thing you wrote on a pitch slide in 2018 because it sounded like where the world was going, and because judges nodded when you said it.

Eight years later I am rebuilding a lab in my own house and learning exactly that. Embedded machine learning. Quantisation. Models small enough to run on hardware you can hold in one hand. The thing eighteen-year-old me put on a slide as a gesture toward the future is now just the thing I am studying on weekends.

No award on the table this time. No pitch. Just the part I always actually loved — making things, understanding why they work, and telling people what I found.

This blog is where that gets written down.