Sorting Out AI Job Titles, Skills, and Career Paths
Looking for a job in artificial intelligence? AI job titles vary wildly, which complicates job searches. Dr. Feiyu Xu, global head of artificial intelligence for SAP Cloud Platform, helps decode AI job titles and explains what skills each position needs.
- By James E. Powell
- August 11, 2020
Upside: What are the primary job titles of today's AI positions?
Dr. Feiyu Xu: Job titles in AI are a mess. AI is a young and rapidly expanding field, so there are no rules or de facto standards yet that prescribe the proper naming of AI jobs. In fact, there are still quite a few more jobs than there are different job titles. The most frequent job titles refer to data, machine learning, or AI in general, such as data scientist, AI expert, ML engineer, ML scientist, AI application engineer, AI research scientist, AI data analyst, and data annotation expert. Other job advertisements search for experts in a subarea of AI such as natural language processing (NLP) expert, computer vision (CV) expert, AI games engineer, AI UX designer, or multimodal UX engineer.
What skills does each position require? Which of these skills can be taught and which are learned on the job?
The most frequently used job titles such as data scientist or AI expert are the least informative. Without reading the full job description, you may not be able to guess what is expected of you and what exactly you will be doing.
In some companies, data scientists are the people who curate, select, prepare, pre-process, and clean data that is then used by the ML experts for training, improving, or evaluating AI models. In other companies, data scientists are responsible for understanding how to solve business problems themselves by exploiting data. They are expected to select the appropriate approaches or tools from traditional data analytics or advanced machine learning and build or train models that can solve the problems.
If AI experts are sought, it could be that the successful candidates are expected to use standard machine learning tools to squeeze out some essential insights from available data. In this case, AI experts are not so different from one type of data scientists. On the other hand, it may be a call for people who build new AI applications involving a variety of AI approaches and algorithms instead of just relying on proven tools.
If you look at a job post, first register whether the job title refers to data, AI, or machine learning as a whole or to a subfield of AI such as NLP, computer vision, VR/AR, games, robotics, etc. Then, notice whether the job type reads scientist, researcher, engineer, architect, or developer. A third indicator may be the seniority level "junior," "senior," or "chief."
Next, forget the title for a moment and carefully read the job description. Usually, the prose is more informative than the title. However, if you are still not sure whether you will be expected to apply tools, develop novel methods, design systems, or build real applications, the job title may give you a hint. A scientist or researcher usually has different success criteria than an engineer or a developer.
If the job ad calls for a data scientist or an ML expert who will apply existing ML tools to solve new problems by analyzing structured data, the applicant is usually not required to have experience in the architecture and coding of software products, but the command of an advanced scripting language (such as Python, R, or Go) is a prerequisite for working quickly and independently. The mathematical understanding of the analytics and ML methods is essential. Still, the math expertise does not have to be deep enough to design new neural architectures or ML paradigms.
If the job involves the development of AI applications, a solid competence in algorithms, software engineering, and systems architectures is required in addition to a good overview of AI. This includes fluency and experience in at least one relevant programming language such as Java, C++, or C#.
Typically, if the position is dedicated to a subfield of AI, at least some knowledge or experience in this subfield is expected. Although it may seem at first glance that all these sub-areas work with similar methods of deep neural learning, the respective properties of the domain data and the particular neural architectures differ quite profoundly.
Most other skills and specializations can be acquired on the job unless they are explicitly requested.
What tips can you offer to a recent graduate to get into the industry? What are the entry-level positions? How long does it take to climb the AI job ladder?
For graduates looking to get into the industry, the good news is that experts have already deemed 2020 as the year enterprises become laser-focused on AI. Although the business landscape has undoubtedly changed alongside current events, the need for enterprises to gain actionable insights faster is more important than ever.
Recent graduates should prefer entry-level jobs in teams composed of people they can learn from. The first couple of years are crucial for completing your education by obtaining expertise and skills that cannot be taught at universities. Sometimes job postings by small or medium enterprises are tempting; they promise lots of freedom and opportunities for shaping AI utilization or AI products. If these positions are not embedded in a capable team, the pressure of coming up with relevant decisions and original contributions may exceed the competence and experience level of professionals just entering the field. The posting company may try to avoid the costs of seasoned experts, and inexperienced applicants can quickly end up in a trap by not being ready to fulfill expectations.
Therefore, strong teams are the best environment for beginners, even if the chances for fast upward mobility turn out to be smaller than in other companies. After two or three years, personal growth will open opportunities that may require a change of unit, site, or employer.
What does it mean to be in the AI industry today, and what do you expect it to be like in two to three years?
The AI industry has experienced a significant transformation over the last few years and has become more commercialized. It's now playing a much more substantial role in the digital transformation of businesses into intelligent enterprises. As businesses continue to digitize, the AI industry is only expected to grow.
It is clear -- artificial intelligence can help enterprises improve their existing products and even create new ones. It is even more critical that AI enables innovative technologies for efficient analysis, planning, and on-demand monitoring of enterprise processes and operations such as supply chain management.
Companies that achieve AI at scale will not only reap competitive benefits but will also support or make significant societal contributions. Just think of advances in new materials design, or understanding and predicting weather patterns.
Of course, pursuing AI at scale comes with essential ethical and social responsibilities. It is up to leaders to foster the educational, regulatory, and ethical environments that enable humanity to advance alongside responsible AI.
What role will SAP play in AI in the next 2-3 years?
At SAP, we're focused on significantly increasing automation levels for our customers by infusing AI into business processes. Our customer's values and business impact are of utmost importance. That's why we combine machine learning, enterprise knowledge from readily available information in ERP systems, human and machine collaboration, hybrid evaluation metrics, and closed development cycles.
We're helping them improve analytics tasks, achieve real-time data analysis, increase accuracy, and remove redundancy from organizations. For example, we have customers using SAP's AI-based Business Document Processing Services, together with intelligent RPA, to automate invoice processing. This solution frees up time for other priority business activities. Additionally, various customers are using SAP HANA and SAP machine learning technology to reduce maintenance costs and increase customer satisfaction by proactively monitoring installations and minimizing outages.
Looking ahead, a key pillar of our AI strategy is to increase the adoption of AI within SAP. To make this a reality, we build on our strong technological foundation (for example, SAP HANA) and reinforce our approach to open ecosystems. To enable the intelligent enterprise, we have been developing and applying the latest AI technologies, such as AutoML, knowledge graphs, and explainable AI -- and will continue to do so in the future.
Through our SAP Business Technology Platform, we are offering customers an expansive portfolio of technology solutions -- including machine learning and AI -- to help redefine and unify business processes for today and tomorrow's intelligent enterprise.
For someone interested in becoming a data scientist, what are the steps they'd need to take to achieve this role? How would this path differ from someone working to become a part of the machine learning industry down the road?
It's easy to confuse the many roles of machine learning engineers, AI experts, and data scientists. For folks looking to break into these industries, it's first essential to understand the differences between the roles, as I mentioned.
In any case, a solid academic foundation is advantageous for all roles -- for example, undergraduate courses in computer science, mathematics, or physics. For graduate studies, it is time to specialize in a field of interest. In principle, it's good to gain experience in a company context, perhaps through an internship or working student activity. There is also an open community in the ML/AI area. Starting your own projects, sharing them with the community, and learning from each other is also a great way to educate yourself.