How to pick the right ML team
(and why most people get it wrong)
I’ve switched ML teams multiple times at Google. Anti-abuse → YouTube Ads → YouTube Shopping Recommendations.
Each move was deliberate.
Most people treat team selection like a job application: they try to impress, they take what they get, and they optimize for brand names on a resume.
That’s backwards.
You’re choosing where to spend 40+ hours a week, which problems will shape your skills, and which managers will define your trajectory for the next 2-3 years. The stakes are higher than most people treat them.
Here’s what I’ve learned.
The fundamental mistake: optimizing for prestige instead of fit
Everyone wants the “hot” team. The one doing LLMs, the one that shipped the paper everyone’s talking about. It’s understandable. But prestige is a lagging indicator.
By the time a team is famous, the hard interesting work is often over, and what’s left is maintenance and scale. You’re joining the aftermath, not the frontier.
The right question isn’t “is this team impressive?” It’s “does this team match where I am right now and where I want to be in 3 years?”
Those are very different questions.
What to research before you talk to anyone