The product already worked, in the sense that Zoom, Drive, and email held it together
ZenClass · GUVI · Product design, shipped with devs
ZenClass already had a solution when I got there — that was the problem. GUVI's bootcamp program ran on a pile of consumer apps duct-taped into a workflow: class links over a messaging app, doubts over email, assignments and material in Google Drive, live sessions on Zoom. Four tools, none talking to each other, each adding a learning curve before anyone learned a thing.
"It works" is a trap
The existing tools were either too generic or too complicated, and either way they leaked. A mentor running multiple batches would get meeting links crossed and walk into confusion about which class was which. Student doubts, sent over email, would sink to the bottom of a thread and disappear. And at the end of a course — the moment that matters most — finding a student's submissions meant digging through folders, for student and mentor both. None of these were dramatic failures, which is exactly why they were dangerous: the friction was absorbed as "how it is."
Knowing when to stop
The real difficulty wasn't the users — it was me. Every pain point suggested a feature, so I kept finding and kept adding, while the developers had moved on to building finalized screens. I was feature-creeping a product the team was trying to ship. What saved me was a fact I almost overlooked: ZenClass batches were already running. So I stopped guessing which features were necessary and started checking against the people living it. "What should we build?" stopped being an argument to win in a meeting and became something I could watch. The live program was the filter.
One platform, two seats
A mentor needs to run classes, keep doubts from getting lost, and review submissions without an archaeological dig. A student needs to show up, ask a question and actually get an answer, submit work, and find their material in one place. The job was to serve both without bloating either side into the other's clutter — one platform that felt purpose-built from whichever seat you sat in.
One place that knows what it's for
The platform is fully built and live, carrying around 200 learners per batch on average, still in use with iterative updates since. The win isn't the headcount — it's that a workflow once smeared across four disconnected apps now lives in one place where a link doesn't get crossed, a doubt doesn't vanish, and a submission is exactly where both people expect it.
Kill features faster
Building against a live program taught me which things mattered — but it also showed me, in hindsight, a few I pushed for that the batches never used. Next time I'd treat my own ideas with the same suspicion I treated the old tools: assume it's unnecessary until the real users prove otherwise.