Is anyone here a recent Type B (Building) Data Science graduate from Type A (Analysis)? Why did you decide to switch?
What changes have you seen in Type B job titles? ML positions are frequently classified under software engineering, and it appears that titles like MLE, Applied Scientist, or AI Engineer are replacing the term “Data Scientist.”
I am aware of this, and I was on a job that began as an A and is currently a B. Jobs that placed me in B responsibilities, required my code to pass engineering reviews, taught me MLOps and CI/CD, and piqued my attention made the transfer easier, I believe. B’s modeling code differs greatly from A’s.
These days, when I’m looking for a job, I just search for “machine learning” and read the descriptions; titles are garbage.
I appreciate you sharing. Do you think the engineering side of the job is essential? To make the change, for instance, should one concentrate more on MLOps and CI/CD or more on algos and modeling?
As a Type B data scientist, CI/CD is probably the most common engineering approach you will employ on a daily basis and is an essential component of the MLOps toolkit. Since your work involves machines and everything that cannot be automated won’t reach production, certain engineering skills are necessary for B roles. Although you won’t be the best engineer, you must be at least competent enough to prevent problems for the team. This is because many Type B scientists bring ML expertise rather than deployment abilities.
Being as full stack as feasible is essential when comparing algorithms to engineering. Solid engineering is necessary for many sophisticated algorithmic applications, and even the most intelligent system will fail if it cannot operate effectively in production.
Usually I’m doing both in my role. Interesting to see them being separated
Usually I’m doing both in my role. Interesting to see them being separated
Same. Small company, Agile development teams on government contracts. I do both the analytical/research phase, then become more of an MLE when implementation needs to happen.