I’m a mid/senior software engineer (~7 YOE) trying to move into applied research in AI/robotics. No PhD

If you’ve hired for research engineer / applied scientist / robotics roles:

What differentiates a strong non-PhD candidate from "Smart SWE who likes ML"?

Which artifacts are most convincing (reproducing papers, open source contributions, working on real-world systems, conference workshop paper, etc.)?

What skill gaps usually show up in interviews?

And if you’ve made the jump, what worked for you?