Main image of article How Data Scientists, Machine Learning Devs Specialize Their Workflows

Data scientists and machine-learning specialists are playing an increasingly integral role in many organizations’ strategy and product development. Are most of these technologists involved in every part of their employers’ data analysis and model training, or do they just specialize in specific areas?

According to SlashData’s Q3 2021 analysis, the answer is the latter: Most data scientists and machine-learning specialists focus on a few parts of the overall data science/machine learning (DS/ML) workflow. The highest percentage is involved in data exploration and analysis; far fewer participate in model deployment, project management, and model health and lifecycle management. Take a look at the breakdown:

The complexity of managing models may account for its relatively small percentage. “However, with more and more models being actively deployed to production environments, we expect that this stage will become ever more important,” suggests the note accompanying SlashData’s data. “The costs of tweaking an existing model for consistent performance are likely much lower than developing a new one from scratch.” (The sample size of SlashData’s survey was 2,412 respondents.)

Complexity is also a possible explanation for why so few data scientists and machine-learning specialists are involved from end-to-end in the DS/ML workflow. “Each stage of the DS/ML workflow is becoming more complex—it’s difficult for developers to stay up-to-date with the newest backend technologies and cutting edge machine learning algorithms, and so we’re seeing fewer people involved end-to-end,” the note adds. “Here, developers are increasing the number of individual stages that they’re involved in, while still remaining somewhat specialized within a section of the overall process.”

For data scientists, machine-learning specialists, and other developers, the lesson here is clear: Specializing in model deployment and optimization can make you an especially valuable player, considering the relatively few technologists who participate in that particular arena. Also, trying to stay as current as possible with every stage of the DL/ML workflow can pay off—companies value technologists who can take a holistic view of critical operations. 

With so few technologists focused on ML model deployment and optimization, specializing in that area can make you especially valuable.