How to grow as a data science professional — introducing the Skill Stack
Professionals need to grow and develop their skills to advance in their career. That’s not different for a data scientist. There are various skills, all contributing to your impact on the project. Developing all would be ideal, but makes it hard to focus. I’ll show how certain skills relate to each other in a Skill Stack, and how to use that to focus your personal development.
Contributions types
A data science professional can contribute to projects in various ways. Often, the core contribution is personal expertise. Hands-on expertise in machine learning, creating models, bringing models to production. Knowing it, and walking the talk. Being a subject matter expert enables the second way of contribution. That of being a guide. It requires some trust, which you can earn partly by showing and sharing expertise. An effective professional will guide his team members, manager, and business through the project. Instead of improving your own efficiency, you now multiply your impact through your peers. Members of other teams and other key decision-makers that you come across in your project will certainly notice that. They are at a bigger distance, as you don’t collaborate with them daily. But still, your expertise can help them in making good decisions. The third contribution is thus to share expertise as advice. So, a professional can contribute by providing expertise, guidance, and advice.
The Skill Stack as model for your skill-set
These different contributions require different skills. Expertise in implementation requires certain skills. To be able to guide the team, you also need coaching and leadership skills. The challenge is how to divide your learning effort among these skills. That division will not be static but needs to adapt to the advancements in your career.
In skill development, the T-shaped skill-set [wiki] is used a lot. The idea is that you specialize in a few topics, and are a generalist in many other topics that relate to your core activities. You could for example be an expert in reinforcement learning, but also know some rough concepts from Bayesian inference. On a more abstract level, you can be an expert in data science, but also know a bit about business development and sales. This perspective provides a rough prioritization of skills, which allows the calibration of attention.
Yet, there are also dependencies between the skills, which is the focus of this article. The skills necessary to contribute through expertise, guidance, and advice, can be grouped in different categories. First, there are technical capabilities; the basis of your expertise. Second, there are personal competencies; behavioral traits that enable and improve your contributions. Third, there are non-technical capabilities, which I refer to as consultancy capabilities. Even if you’re not a consultant, a data scientist can still utilize consultancy skills. Together they result in the Skill Stack:
The skill stack shows how the types of skills depend on each other. The consultancy capabilities depend on technical capabilities. A professional can only be a consultant in data science if he or she knows enough about data science. Moreover, if you are going to give presentations, public talks, or board room advice, you also need some personal competencies. In this case, in communication. The advantage of separating these three is we can now focus on relevant improvements. Furthermore, each kind of skill requires a different approach to realize growth.
Skill-set 1: Technical capabilities
The technical capabilities are the skills related to hands-on work, actually developing a data science product. Development in Python, writing a model fit pipeline, designing a deep learning network architecture, deployment using Terraform, and so on. These are the individual contributions. Every capability has its own specific development path. Together, they shape your expertise.
Initially you need to learn a lot of new capabilities, then you apply them, gain more experience, and later you might teach those skills to others. The initial learning phase can be separated into three aspects: gaining awareness, learning theory, and practice. You need to be aware of the aspects of a task (awareness) to find resources to read (theory). But theory on its own is not enough. Practice is necessary to be able to discover unexpected situations, and to discover the devil in the details. Before we dive deeper into ways to improve the capabilities, let’s first cover the other two types of skills.
Skill-set 2: Consultancy capabilities
The consultancy capabilities are the non-technical capabilities, applied to create business impact. They relate to tasks or events like board meetings, creating roadmaps, and planning. They are more related to organizational skills, teamwork, and personal efficiency. Initially, you need a set of technical capabilities to be able to create output. The consultancy capabilities then enable to leverage that output into business-impact. That impact develops in three phases.
First, you’ll need effective execution. You develop a habit of personal efficiency, applying, for example, Getting Things Done or eXtreme Programming. Effective execution results in a high-quality implementation. After learning how to manage yourself, the second phase is to optimize the solution. The perspective grows a bit wider. You’re not solely focusing on your own tasks and goals, but the goal of the whole project. The effective solution focuses on the right solution, for the right problem. This requires more business understanding and more work in analyzing the business. When that understanding grows and is combined with more experience, you can create impact on an even larger scale. Your impact then focuses on an effective strategy. While the focus on an effective solution ensures you solve the right problem in the right manner, the focus of an effective strategy improves the kind of problems and the direction of the organization. Thus, in steps of spreading impact, we see effective execution, effective solution, and effective strategy.
Support of consultancy capabilities by technical capabilities
To be able to execute the effective execution, solution, or strategy phase, you need underlying technical capabilities. One needs to have technical skills to be able to actually execute your tasks. You also need sufficient knowledge to be able to evaluate alternative solutions and discover opportunities. When you start guiding the project team, you can only do so when you have the required knowledge to provide valuable guidance. When transitioning further to strategic advice, one can make an argument that detailed technical knowledge is not required anymore. You, however, do require a good grasp of concepts, complexities, and implications of technical decisions. Usually, you only learn these when growing as a technical individual contributor and gaining experience. With the resulting general knowledge, you are able to provide truly valuable advice.
Skill-set 3: Behavioral competencies
To be able to realize that impact, you also need behavioral competencies. For example in communication, networking, politics, and coaching. These can be regarded as soft skills, and are more tied to you as a person, than to the current role in your project. Such competencies are often not utilized independently. They enable and support the other, more visible, actions you do, especially the consultancy capabilities. Developing them often goes in sync with the development of those consultancy skills. Trying to change the direction of your project or client’s business, requires various competencies, allowing you to develop those too. It is, therefore, more difficult to develop a certain competency independently. In my experience, it helps to focus on a consultancy goal, while being aware of the underlying competencies. Then you can find ways to improve and adjust them, to become a more effective consultant.
Take for example presentation competencies. Giving some great presentations can help share your vision, and steer the direction of the project you’re working on. It can be really beneficial to know the goal you want achieve, in this case the direction of the project. And subsequently, know which capabilities might help. Thus practicing your presentation thoroughly, not only to save this project, but to improve your presentation capability. Which in the future, might also benefit other projects.
There are a lot of behavioral competencies that can be developed. They can be grouped as relating to the following items:
- Personality
- Entrepreneurship
- Management
- Motivation
- Analysis and decision-making
- Communication
Support of consultancy capabilities by behavioral competencies
The consultancy capabilities are empowered by underlying technical capabilities and behavioral competencies. The technical capabilities provide the knowledge and experience. The behavioral competencies provide the skills.
Each stage of consultancy focus requires different behavioral competencies. At effective execution, you’ll benefit from work ethic related competencies like discipline. A focus on an effective solution will require more organizational competencies like task management. When you focus on an effective strategy, you’ll also need competencies related to spreading your vision.
So we can draw a general relationship between the specific consultancy phase and certain competencies. These dependencies are not black or white, but rather an indication of major relevancy.
Each competency will have some benefit, independent of the consultancy phase. However, when transitioning from phase to phase, different competencies will become more critical.
When focusing on effective execution, the initial goal is to implement the chosen solution. The project is defined, the direction is clear, we need to execute it. At this level, relevant competencies are:
Customer focus, Listening, Sociability, Verbal communication, Adjustment, Effort, Integrity, Discipline, Cooperation, Flexibility, Analysis
Still focusing on effective execution, the next level would be to adjust the choices of the solution. We think along with the customer and influence the plans. The newly relevant competencies are:
Planning, Organization, Initiative, Self-Development, Ambition, Learning ability, Organization sensitivity, Results-orientation, Persuasiveness, Tenacity, Quality focus
The subsequent focus is on an effective solution. You not only determine how to implement the solution, but also determine what solution is needed. You thus change the course of the project. This allows for new and different solutions, even approaches not considered previously. To be able to effectively operate with this freedom, the consultant needs to help define the direction. This requires new competencies like:
Creativity, Independence, Negotiation, Stress resistance, Impact, Progress monitoring, Task management
If we transition our focus to more strategic decisions, we need to provide advice that influences multiple projects. That will require the following additional competencies:
Managing, Coaching, Delegation, Vision, Decisiveness
If the activities focus more on advice than leading, and the advice focuses more on the business than on the technology, also competencies like the following will help:
Entrepreneurship, Market awareness, Networking
Continuous improvement in the Skill Stack
To recap, a consultant is able to have an impact, through supporting technical capabilities and behavioral competencies. This enables the consultant to have an impact on the individual level, on the solution, and on the strategy. To improve on these skills, you can follow the Plan, Do, Check, Act cycle of development [wiki]. Use the Skill Stack to identify your learning focus. Be aware of possible improvements and make sure to change your course of action accordingly. Verify the changes, and act on the received feedback.
Dreyfus model of skill acquisition
But, what are the progressions in the individual skills? This brings us to the Dreyfus model of skill acquisition [wiki, figure]. The Dreyfus model distinguishes 5 stages:
- Novice; needing detailed recipes for execution
- Advanced beginner; detecting nuance in rules
- Competent; learning organizing principles
- Proficient; recognizing patterns, following standard approaches
- Expert; just knows what to do
While the previous part shows how the different skills intertwine, the Dreyfus model shows the path along which to develop them. It might be most intuitive for technical skills, but it is not limited to those skills. The Dreyfus levels can actually be applied to each individual technical capability, consultancy capability, and behavioral competency.
It would thus be tempting to align the different skill-sets in the Skill Stack (technical capabilities, behavioral competencies, and consultancy focus) with the Dreyfus levels. The Dreyfus levels are however applicable to specific skills. Skills like modeling in SK-learn, presenting, and board meetings. While the stack shows progressions, the progressions are in focus and tasks, not per skill level. In practice, one might summarize different skill levels in a general statement. Like a statement that someone is a ‘proficient machine learning engineer’. But note, that person might still be a beginner in a certain specific technique.
Goals as guidance for skill acquisition
Knowing how the skills intertwine and how you develop them, it’s time to focus. Goals can provide that focus and guidance during your personal development. As an example, let’s assume you want to change the course of your project. That is a goal which you can use as guide. Given that goal, find the relevant skills you need to achieve it. Those skills are the ones to focus on along the way. To ensure you’re focusing on the right skills, find technical, consultancy, and behavioral skills relevant to the goal. With a clear idea about the goal and relevant skills, you can continuously improve yourself using the Plan-Do-Check-Act cycle. Strive for personal excellence and keep growing. Besides a guide for focus, the goal also provides an higher purpose, giving more meaning to the skill acquisition. Which will in turn improve motivation.
For each found skill, be honest about your current (Dreyfus) level. Being honest can be hard. It’s easy to overestimate how much you know, and underestimate how much more there is to know. That is the Dunning-Kruger Effect. In short, your confidence will grow faster than your knowledge, until you discover how much more there is to grasp. Don’t evaluate how much you already know or compare yourself to other beginners. Look at how much guidance you need, how much examples you look up, how often you get unexpected and unintended results, and how well you can explain details to others.
Based on your actual level, determine the next step. As a novice, it’s probably best to search for tutorials and guidelines. To become an advanced beginner, determine the the context of the rules. To become competent, understand the why behind the rules. Always practice, to really experience the situations. Notice if the observed result was expected, or not, and learn from it.
Environment as enabler for skill acquisition
To be able to achieve your goals, there are three key ingredients. First, your environment needs to have opportunities, allowing for personal development. Second, you’ll need to define your personal ideal direction, within all those opportunities. And lastly, the environment needs to provide feedback, ensuring that you learn fast and you’re able to adjust your direction.
Your environment should allow you to create and take opportunities. Your working environment is your playground. It should allow the discovery of new skills and to stumble while trying to achieve new goals. The ideal environment allows time to be spent on tasks you don’t fully master yet.
There should also be a variety of projects. Projects that require new skills, triggering additional growth. With a variety of projects you’ll get a broader experience. It’ll also become easier to discover which kind of career direction suits you best.
We already mentioned you need goals to guide you. Ideally you get some guidance to determine good goals and to correct your course while achieving them. Guidance can be brought by coaches or managers. But it can be brought best by mentors. Mentors who have walked your path in the past. They can be your role model, preventing you from making the same mistakes. Furthermore, they can be an example of what you don’t want to be. They’ll have gathered a lot of knowledge, so pick their brains. They are often more than happy to improve your future.
Lastly, you need feedback for adjustment. Feedback can come from more senior colleagues or mentors. But it is sometimes hard for them to understand your situation and specific struggles. Probably the most valuable feedback is received from peers. Peers, like your colleagues, understand your situation. They can give you feedback, can explore new techniques with you, and allow for knowledge sharing. Sharing what you know is the best way for them to learn, and for you to get feedback on your knowledge. They also provide a team, giving you a sense of belonging. They are all on similar paths, which gives you different perspectives, challenging your chosen direction. When trying to understand new things, discussions with them are a good way to find out how things work. In short, peers are a great way to exercise learning and understanding.
Your personal best environment
Finding the environment that matches your preferences will be a personal quest. And be aware, that you can also adjust your existing environment to make it match.
I found my spot at BigData Republic, a small consultancy firm in the Netherlands. We like to regard ourselves as a boutique of specialists. Great smart colleagues, there’s always a topic that someone else knows more about than you. As a consultancy firm, we have the opportunity to provide a variety of projects to our consultants. As knowledge growth is a top priority, we make knowledge sharing also a top priority. Every Friday we are at our own office, instead of at the client’s office. This provides an informal way to share knowledge and experiences. Slightly more formal, but still easy, are our weekly chapter sessions. During these sessions we discuss projects and new techniques, we read articles and have brainstorm sessions. We also teach each other new techniques using more formal dedicated sessions. Those dedicated sessions are workshops, prepared by our own colleagues, where we dive deep into various technologies. These initiatives make knowledge sharing part of our DNA. We challenge ourselves by taking our knowledge sharing a level up. Creating formal in-house training for clients is the best way to challenge our own expertise.
This approach has worked very well for us and has created a great environment. We all wished we had such an environment at the start of our careers. Therefore, we created a new sister company: Vantage AI. The same setup, but for colleagues who are at the beginning of their data science career. We have office days on Friday, consultancy projects, training, and mentorship from BigData Republic. A difference is that we introduced groups. New team members join as a group and get the training together. This enhances the interpersonal relationships and availability of peers.
Have a nice trip
Whether you’re at the start of your career or not, I hope the Skill Stack can help you adjust the course of your career to a better direction. With the Skill Stack as mental model, goals as guidance, and a good environment, you can set yourself up for success.
During my years at BigData Republic, I’ve tried to create a place where my colleagues and I can grow together. This has resulted in the environment described above. While I’m coaching some data scientists, I had to discover a way to do so along the way. The search for knowledge and trying to comprehend the different aspects, resulted in this new mental model, The Skill Stack. It helped me in my own growth and while coaching colleagues, hopefully it helps you to. Either way, I’d be glad to hear your experiences from you.
About the author
Gerben Oostra is our lead machine learning engineer at BigData Republic, a data science consultancy company in the Netherlands. He has been with us since the very beginning, allowing him to shape the environment. Together with the team, he created the ideal environment — as described above — for his colleagues and himself.
If the described environment sounds like a good spot for you too, either at BigData Republic, or at Vantage AI, please let us know.
We hire the best of the best in data science, data engineering and cloud engineering. And whoever joins, will have a lot of opportunities to improve their skills even further. If you are interested in using machine learning techniques on practical use cases, feel free to contact us at info@bigdatarepublic.nl.
