Puffed-up title or truly different skills?
In the legacy enterprise realm this data engineer/data scientist split would be akin to the data-management types (DBAs and ETL/data-integration professionals) versus the data analysts and analytics professionals (handling BI and data mining, respectively).
I've seen some cases of people just upgrading their titles without upgrading their skills. But people who have truly graduated to data engineer can manage Hadoop clusters, handle data processing on the platform, and identify, move and, perhaps, cleanse and normalize subsets of data of interest for deeper analysis by the data scientists. Data scientists, meanwhile, can write algos and develop data-driven applications from scratch whereas old-school data miners are more likely familiar with SAS, SPSS and perhaps R-based algorithms that can be called and tested in building models on supported workbenches or studios.
The old skills are still in demand, but the new skills are much rarer and sought after by pioneering data-driven organizations exploiting varied and high-scale data types not stored in old-school data warehouses.