A Quick Look at Recent Trends in ML Engineer Job Requirements
I skimmed through ML engineer job postings over the past month, and it feels like the commonly required stack has changed.
Previously, having experience with model training frameworks was enough, but now quite a few places treat MLOps / pipeline building experience as a baseline and even ask for LLM application experience like RAG.
Also, Docker and Kubernetes are now often listed not as options but as basic requirements. It made me think that people who only studied theory in grad school would have a tough time with job hunting these days.
Of course, those who are really good can go anywhere, but for mid-level developers like me, it's a bit of a scary trend—if we don't keep learning, we'll soon fall behind.
5 answers
Agreed, MLOps is the standard these days.
I've been feeling this too while preparing for a job change recently—without a pipeline, it's really hard to even get past the resume screening. Docker you can just pick up, but Kubernetes takes a bit of time. Still, even learning Airflow through a side project will pay off a lot down the road.
Well, aren't there still plenty of places that value modeling itself? It's more that startups and some big companies emphasize MLOps, while research roles have different criteria. But yeah, it does seem like the requirements have definitely increased compared to before.
I feel the same wayㅠㅠ Technology changes so fast it's hard to keep up.
As someone who only did theory in grad school, I completely relate. Writing papers and working in practice are truly different worlds. These days, MLOps engineering at least has broader paths forward, so I think it's important not to give up and to learn things one by one.