I've been in the data science space for over a decade, and it feels like the "full-stack unicorn" role from the 2010s is gone. The core tasks seem to have split into two distinct paths:

The Analyst/Strategist (Decision Scientist): Focused on causal inference, experimentation, and business strategy, who uses AI as a powerful analysis tool.

The Builder (AI/Data Systems Engineer): Focused on the production architecture of it all—building the data pipelines (Kafka, Flink), MLOps infrastructure (Kubernetes), and now the agentic workflows (LangChain, RAG).

It seems the middle ground—the generalist who could do a bit of everything but wasn't a deep expert in production engineering or causal stats—is becoming obsolete. The "science" is now either in the rigor of experimentation or the engineering of complex, non-deterministic systems.

For those of you who have been in the field for a while, does this model resonate with what you're seeing?

Are you experiencing this bifurcation in your own companies or careers?