When to Say No to “Platform Work”

I've watched too many talented senior data engineers quit over the years. Not because of salary, or work-life balance, but because they spent years building internal tools that nobody used.

This is the platform work trap, and if you're a data or ML engineer at a product company, you've probably felt it. That subtle pressure to stop building things that matter to the business and start building "infrastructure" instead.

Sometimes platform work is exactly what you should be doing, but most of the time it's a career dead-end wrapped as a promotion.

Here's how to tell the difference.

Platform work is anything you build primarily for other internal teams to build on.

Not "infra" in the boring sense (like provisioning VMs, patching kernels), and not "shared code" in the normal sense (like a util function, a library). Platform work is when you're trying to create a paved road: