The Rise of the Industrial Data Scientist in an Industrial AI World
The industrial sector is currently in the middle of its next pivotal transformation, which this time it is a digital transformation driven by the industrial internet of things, AI and machine learning (ML) algorithms. Compounding it are generational shifts in the workforce between veteran domain workers and newer hires who are more tech-savvy but lack the same level of operational expertise.
At the heart of this transformation is the industrial data scientist, a new breed of tech-driven, data-empowered domain experts with access to more industrial data than ever before, as well as the accessible AI/ML and analytics tools needed to translate that information into actionable intelligence across the enterprise. Industrial data scientists represent a new kind of crossroads between our traditional understanding of citizen data scientists and industrial domain experts: workers who possess the domain expertise of the latter but are increasingly shifting over to the data realm occupied by the former. In fact, based on a recent survey, nearly 40 percent of data scientists in the industrial sector today come to the job with a background in chemical engineering, process engineering or industrial engineering and not in computer science or software engineering.
Amid these workforce shifts, digitalization trends and transformations of traditional industries, industrial data scientists can play important roles. Their emergence can align with and complement a similar rising tide of AI and ML adoption in industrial organizations.
The role of the Industrial Data Scientist
As chemical engineers who work in and with data science, industrial data scientists focus on solving real-world problems in the field. They draw on their domain experience to incorporate domain knowledge into data science projects – a level of expertise that traditional data scientists do not naturally carry. At the same time, industrial data scientists help to demystify and share the value of data science across organizations. Their dual position as domain experts and data scientists serves to evangelize the work of data science within the organization that might otherwise get lost in translation. With the domain knowledge they bring to the table, industrial data scientists help create the potential for new business opportunities while also having the skills to translate raw data into actionable insights.
By contrast, more traditional data scientists, particularly those with researcher profiles, leverage their academic experience to tackle complex problems. Their role is typically more focused on long-term projects, such as developing sophisticated new ML models to solve those problems, improving key performance indicators and generating new evaluation metrics. While this will remain a crucial role for industrial organizations, digitalization initiatives and workforce changes will shift more domain experts from traditional engineering positions toward data-driven work. As that happens, the industrial data scientist is poised to become an increasingly integral part of how industrial organizations digitally transform and leverage industrial AI applications to drive new value.
The Challenges Holding Back Industrial Data Scientists
An industrial data scientist’s core mission is to build more comprehensive, high-performing and sustainable AI and ML models that address focused, real-world use cases. These industrial AI models span the asset lifecycle and are pervasively leveraged to guide industrial organizations through digital transformation to maximize productivity, efficiency and production outputs – while delivering on the vision of the self-optimizing plant. But this mission is often impeded by organizational, technical and process challenges that prevent industrial data scientists from being able to do what they do best.
Some of these challenges include:
Coordinating domain knowledge resources across subject matter experts and managing pertinent data across scattered files and disparate tools.
Collaboration with other domain experts to tune, test, train and improve models on an ongoing basis to deliver on business goals.
Determining the right set of libraries and AI/ML environments to use.
Figuring out where to put their ML code and how to version and collaborate on that code.
Handling and scaling additional resources as necessary due to increased computational complexity or data volume.
Connecting to diverse data sources and overcoming data integration and mobility hurdles associated with accessing real-time historical data.
Sharing results and deploying models in production.
Pulling together the IT, data engineer, DevOps and integration efforts needed to productize proofs of concept.
Empowering Industrial Data Scientists
Industrial data scientists should not have to figure this all out themselves in to build scalable, monitored and secure Industrial AI models.
The value of providing tools that enhance the work of industrial data scientists, which can free them up to do what they do best, speaks for itself.
That includes faster time to market, which is crucial in a VUCA (volatile, uncertain, complex, ambiguous) environment, enabling a company to be more agile, beat its competition to the market and secure market share as a first mover.
Another benefit is increased productivity. Data scientists are expensive and automating tedious parts of their work yields significant ROI.
Stronger innovation is another driver. Companies so often succumb to the “valley of death” between the idea phase and productization of an innovative idea, and this gap is exacerbated by the challenges data scientists face. Enhancing their work with a mature industrial AI environment helps ensure improved concept testing and improves the likelihood that a good idea gets implemented into a product.
To supplement the efforts of industrial data scientists, industrial organizations must invest in a robust, scalable and secure industrial AI infrastructure that creates an environment where data scientists are free to develop, train, maintain, deploy and execute industrial AI and ML models, without any of the usual roadblocks. This means creating a full-stack approach that abstracts complexity in data science projects, makes it easier to fetch necessary data, facilitates collaboration among domain experts like process engineers, and enables seamless deployment of AI and ML projects into production.
About the Author
Bill Scudder is a senior vice president and general manager for artificial intelligence of things (AIoT) products at AspenTech. He previously served as the company’s senior vice president and CIO and remains responsible for the company’s IT organization. Scudder has more than 25 years of IT leadership experience, developing and implementing mission-critical, global technologies and building the operations and IT organizations to support them. He previously worked for Sonus Networks, EMC and for Smith & Nephew. Scudder holds a bachelor’s degree in mechanical Engineering from Rensselaer Polytechnic Institute, and an MBA from Boston University.