Building ML Teams: What to Look For

Building an ML team is harder than it looks. The talent market is competitive, and the skills needed vary by problem. Here's what we've learned about hiring.

Data scientists aren't all the same. Some are strong on statistics and experimentation. Others excel at engineering and deployment. Some are domain experts who know ML. Know what you need before you start interviewing.

The best predictor of success is actual project experience. Not certifications, not degrees—real projects with real data. Ask candidates to walk through a project end to end. Where did the data come from? How was it cleaned? What was tried? What worked?

Communication skills matter more than in traditional software. ML practitioners need to explain complex concepts to non-technical stakeholders. They need to understand business problems and translate them into technical approaches. Pure technical skills aren't enough.

For early teams, generalists beat specialists. You need people who can handle data engineering, modeling, and deployment. Specialists become important later as the team grows.

Don't over-index on academic credentials. Some of the best ML engineers come from non-traditional backgrounds. A physics PhD might be great, or might struggle with messy business data. Look at what they've done, not where they studied.

Finally, consider build vs. buy for the team itself. External help can accelerate early projects and train your internal team. But you need internal expertise eventually—vendors can't own your ML capability forever.