I see the typical "Data Scientist" role responsibilities getting divided between Data Engineer, Machine Learning Engineer, and MLOps engineer roles very soon. You can no longer know just a few Python libraries and call yourself a Data Scientist. Domain knowledge and the software around machine learning will take (and are already taking) precedence.

Of course ML research will continue but it'll be limited to the core research teams, which I believe many organizations won't see any value in establishing. Moreover, how many of us data scientists code algorithms from stratch if there's an open source implementation available? In my opinion, if you want to continue working as close as possible to the Data Scientist path, learn to be better at data engineering and applied research rather than trying to reinvent the wheel.