About — Steve Tamekuni

← Back

I found product design in the last class of my undergraduate degree at UC San Diego. It was called design engineering and it was really a product design sprint. That was enough to send me to the master's program at San Jose State, and I've spent the eighteen years since picking up the hard and soft skills the work actually takes.

Most of that time has been on internal tools and enterprise platforms — the software people have to use rather than choose. Hiring products at LinkedIn, workforce management at Zoom, infrastructure capacity at Meta. The through-line is systems with competing needs, where the person downstream of a decision is rarely the person making it. The work I'm proudest of is where I set the standard the team designed against, or built the framework that produced the proposals, rather than executing a brief someone else had already scoped.

My last role at Meta is where AI became part of how I actually work. I use Claude and Claude Code daily — to understand engineering constraints, to read a codebase before I design against it, and to build in code that engineering reviews before it ships. It lets me carry a question further before I hand it off, and reach discovery that would otherwise wait a week. There's still more to learn than I expected.

What I'm drawn to

Deciding what a system should stand for

Enterprise work usually arrives as a list of features rather than a problem. At Zoom I was the only designer across workforce management for ten months, and it arrived that way — notes on activities, an HR integration, a rotational template. It only cohered once I stopped treating those as separate features and named the standard underneath them: a system that acts on people at scale owes them, and the person responsible for them, an account of what it did. That standard is what made a backlog into a product position, and it's what I argued from in every review after.

Protective design

The happy path is the easy part. I spend my time on what happens after — the error state, the empty state, the decision someone inherits three steps later without context. At Meta, a capacity recommendation was accurate and still ignored, because the explanation for it lived somewhere other than where the decision got made.

AI that earns its place

AI is worth adding when it removes friction from work someone is already doing, not when it invents a new thing to learn. The test I apply is whether it shortens the path to a decision the person was going to make anyway. If it doesn't, it's a feature looking for a reason.