20 Soft Skills to Look for in Candidates for your Machine Learning Team

Published in

Towards Data Science

9 min readFeb 16, 2023

While education, skills, and experience provide the technical foundation and are essential for a capable machine learning (ML) team, the team will only turn into a strong and successful one when combined with the right soft skills.

Image Reading Robot ML generated using https://creator.nightcafe.studio

Education, skills, and experience are very important traits for a capable ML expert. Having a solid educational background in mathematics, computer science, the natural sciences, and statistics provides a strong foundation for understanding the underlying theories and algorithms that drive ML models. In addition, acquiring practical skills through hands-on experience with various programming languages, libraries, and tools is critical for implementing and deploying successful ML solutions. [R1, R2]

However, the right set of soft skills can transform a team of capable ML experts and turn them into a successful one. These skills complement technical skills. [R3] Soft skills are personal attributes and abilities. In the field of ML, soft skills are especially important as they enable experts to work effectively with stakeholders, clients, and other team members to ensure that the right problems are being solved, and the right solutions are being developed and deployed.

Get ready to level up your team! In the next 20 points, I’ll share my personal take on these abilities. So let’s dive in and discover how to bring out the best in ourselves and our teams!

Image of reading robot ML generated using openAI dall-e

Working with purpose: It is important to have clarity about the purpose of an activity. A clear view of the target allows working consistently. In ML projects, it is very easy to go from one interesting analysis to the next. The risk is that you end up somewhere without finding a solution to a real-world problem. Working with purpose can help develop meaningful solutions. [R4]
Working with discipline and focus: Working with self-discipline is a major contributor to success. It takes self-discipline to correct past behaviour and not repeat mistakes. It makes it possible to develop good work habits and maintain adequate quality standards. [R5] In the modern workplace, distractions are everywhere. Focusing on a specific purpose and keeping your full attention makes it possible to achieve quality results within a finite amount of time. [R6]
Intellectual rigour and flexibility: When developing ML solutions, it is important to be clear about the underlying assumptions, apply logical and rigorous reasoning, and come to a conclusion. At the same time, it is also important to preserve the mental flexibility to re-question assumptions and revisit conclusions if the results look suspicious. In my experience, ML-generated results that look too good to be true are often not.
Time management: It is difficult to find time for focused work. The team can introduce dedicated time for focused work, but this can only enable concentrated work if each team member makes the most of the time available. Another aspect of time management is the ability to work consistently and set clear priorities to manage interdependencies between projects and teams and meet deadlines. Remember: you will never finish all you want so make sure you prioritise! [R7, R8]
Cross-cultural competence: ML experts can come from a wide range of educational backgrounds, nationalities, and age groups. It is important that each team member is able to deal with cultural differences. For example, in our team, English is predominantly spoken, but for most of us, English is a foreign language. It is important to listen carefully and ask questions until both sides of the conversation are sure they have reached a common understanding.
Life-long learning attitude: The field of ML is wide and the demands and expectations on the team are diverse. A deep-rooted learning attitude with a general interest in field-specific topics, and content that is loosely related to it, as well as other areas where interesting problems are solved, can become a rich source of inspiration for problem-solving. This basic attitude can form the basis for teams to quickly become familiar with the subject-specific conditions in new projects. Incidentally, lunch break discussions also get a lot more fun and interesting! [R9]
Frustration tolerance: When developing ML solutions, experimenting and testing new methods with unknown outcomes are a part of daily work. ML experts have to take the risk that comes with entering unknown territory. Things should not go wrong, but they will, e.g. the data is insufficient (quantity, information content, quality), the algorithm may not converge, and many more. The team needs to be able to stick with a problem when it gets difficult, be able to cope with setbacks and keep going anyway. It won’t always work the first or second, or third time — ML is hard, and we have to deal with it.
Sense of responsibility: Working with data means working with trust. This comes with a great deal of responsibility: For the data, the team is working with, for the outcome of the algorithm applied, and for the unintended results that arise for example from bias in the data. It is of utmost importance that ML teams acknowledge and are not afraid to take this responsibility. [R10]
Sense of accountability & ownership: The development and operation of ML solutions are complex and often mean that the team has to deal with uncertainties. It is important that everyone in the team takes ownership of their respective work. Teams with a healthy ownership mindset often have a culture of “if I break it, I fix it”. However, ownership requires accountability. Each individual has to be accountable for him or herself and must have the space and safety to be honest about the results. [R11]
Engineering mindset: A ML model regardless of its accuracy is of little value, if it can not be deployed, applied to real-world data, scaled, and maintained. A successful team develops end-to-end solutions that are maintainable, scalable, and robust. It comes with two aspects: First, the data — data is often messy and need a lot of work and care to unveil its valuable information content. Second, the software (& cloud) engineering. Here the software engineering best practices apply. [R12]
Analytical and critical thinking: Don’t trust the machines! ML algorithms after all solve a numerical optimization problem. They take numbers as input and will output numbers. Whether the outcome is reasonable or solves the given problem, needs to be checked by a team of experienced ML experts.
Collaboration: Trust and safety are the foundation of any true collaboration. Each team member must have trust in their own competencies and their limitations, in order to build trust within the team and work on continuous improvements. Safety in this context means: first, within the team, we respect ownership, but actively contribute to the success of the ML product. Second, everyone makes mistakes each team member feels safe to own their mistakes, correct them and learn from them. This enables each team member to be a strong and reliable collaborator.
Role-based mentoring: Role-based mentoring can be an effective way to foster personal and professional growth. There are two facets to this approach: firstly, seeking out mentors either within or outside of the team in order to engage in active learning, receive guidance, and exchange ideas. Secondly, seeking out mentees to provide guidance, e.g. in form of code reviews, discussion of data exploration or to challenge ML solutions.
Problem-solving: Every request we receive represents a problem that someone has not yet been able to solve. Our job is to explore whether there is a potential solution based on the available data, which is an intrinsically difficult challenge. To tackle it, everyone on the team needs to have their personal toolbox and be able to approach, untangle, and ultimately solve the given problem. This often involves applying advanced analytics to complex data sets, developing effective algorithms, and finding innovative solutions. This capability enables a valuable contribution to the team, the company, and its customers.
Effective communication: Both the active speaker and the active listener share the responsibility for good communication. The speaker needs to be able to articulate complex technical concepts and results when communicating with collaborators, stakeholders, or other experts. [R13] In ML development clear resource and time planning can be difficult. It is often called Data Science for a reason. Therefore, the skill of negotiating resources and deadlines is an important aspect of the team's work. [R14] Often, the data cannot live up to the hopes of stakeholders which can lead to friction. Equally important, the nature of teams with cultural and educational diversity, different competencies and at times conflicting goals can lead to truly difficult conversations. It needs patience, openness to understanding, and a healthy amount of empathy to handle such conversations well. [R15] We listed here the three major aspects of strong communication necessary for a successful ML Expert.
Coping with ambiguity: Ambiguities arise from competing ideas, unclear outcome vision, conflicting interests, and limited information. The skill to reason and adapt plans based on available information is crucial to reach a conclusion and determining the best next step. [R16]
Thinking strategically: The capacity to envision the overall solution and its impact on the team, organization, customers, and society are valuable skills for an ML expert. This competency, combined with a deep understanding of the interconnected intricacies of product development in ML, empowers them to stay focused on the big picture, anticipate obstacles, and think several steps ahead. Consequently, the expert can communicate with stakeholders and customers with clarity and prioritize the most critical areas for success.
Organisational skills: Typical challenges in ML product development include intricate interdependencies, unforeseeable obstacles, and incomplete information, such as uncertainty around the adequacy of available data to address the problem. It is a key skill to plan what can be planned, deal with the unexpected, set priorities, allocate the right resources, and deliver results effectively. [R17]
Business acumen: It is the ability to identify and prioritize the right decisions that positively influence the economic success of a company. An important prerequisite is to understand the business problem and the customer needs. Then it is a challenge to realize them effectively and in a technically performant way. The performance of the technical solution relates to the quality of the ML model, and, equally important, to its cost-effective implementation. Strong business acumen enables the ML expert to contribute to the company’s profits.
Working with a customer focus: Give the customer what he needs, not what he asks for.
It is an important skill to use knowledge and competencies in a way that helps the customer in the end. The client should have confidence in the ML product and the development team. Customer focus means understanding the customer’s needs and developing an adequate solution. [R18]

“Our plan is to lead the public with new products rather than ask them what kind of products they want. The public does not know what is possible, but we do.” Akio Morita (1921–1999) was a Japanese businessman and co-founder, CEO, and chairman of Sony.

20 Soft Skills to Look for in Candidates for your Machine Learning Team

Recommended Reading List

Written by Lydia Nemec