This short paper states a scientific position that proposes a new vision of data science pipelines defined as queries, namely data science queries (DSQ's). Different from classic queries, the results of DSQ's are not only data but also estimated models with associated error and performance scores. Besides, queries can have different attainable results according to the algorithms that implement them behind the scenes. A data scientist must choose the best or most adapted result according to given expectations related to a target domain. In this sense, it is possible to consider DSQ's as fuzzy queries that estimate results and choose those close to a combination of expected criteria. The paper discusses the aspects to consider for modelling a data science pipelines as fuzzy queries and possible research directions.