After a decade in data science, I’ve noticed the decline of the "full-stack unicorn" role from the 2010s. Key responsibilities have now diverged into two main paths:
-
Analyst/Strategist (Decision Scientist): This role emphasizes causal inference, experimentation, and business strategy, utilizing AI as a key analytical tool.
- Builder (AI/Data Systems Engineer): This role focuses on production architecture, creating data pipelines (like Kafka and Flink), MLOps infrastructure (such as Kubernetes), and implementing agentic workflows (including LangChain and RAG).
The once common generalist who could handle various tasks but lacked deep expertise is becoming less relevant. The essence of "science" now lies either in rigorous experimentation or in engineering complex, non-deterministic systems. This observation raises the question: Are others in the field witnessing this bifurcation within their companies or careers?
