How to Keep on Developing as a Data Scientist
While I know it might sound like a cliché, being a data scientist often involves having the mentality of a lifelong learner. The field is developing so quickly that it takes time and lots of effort to stay up to date with the latest developments, whether it's the state-of-the-art ML model, a new data manipulation library, or a just-released arXiv paper you'd love to implement. No wonder so many of us (myself included) suffer from imposter syndrome.
And while nowadays there are many opportunities to learn, our most precious resource is time. We can't (or at least shouldn't) spend most of our waking hours working and learning, as we would risk burning out quite quickly. So that's why in this article, I would like to focus on the development possibilities during your 9 to 5 (or any other time range that applies to you).
I know that each company is different, and there is a high chance that you might already employ some of the things I mentioned here. And that's great! From my side, I'd consider this a success if you find at least one new idea on how to keep learning.
Tips on how to grow as a data scientist
The tips I will share in this article are based on my own experience so far and on things that I have heard from my friends and colleagues working in the industry. It is likely that some of these might be difficult to pull off in some companies. But I do believe that it always makes sense to try and you (and your managers) might be really surprised about the positive outcome!
Maximize the learning while working on your regular projects
While this tip might not be anything groundbreaking, you will hopefully continue to learn a lot while working on your regular projects at work. After all, that will be the majority of your workweeks. My advice here would be to make the best use of that time. Here are some ideas:
- Try exploring new ideas for your projects, such as using new models, approaches, or tools.
- Research what other companies are doing to solve similar problems.
- Absorb domain knowledge from your business colleagues. While not immediately obvious, this often leads to bigger improvements in your projects than using the latest model or spending hours tuning hyperparameters.
- Don't neglect non-DS related skills, but more about that below.
Plan your development
Again, this shouldn't be anything new, but it can often be overlooked. We regularly tend to focus on the short term and get into task completion mode. We tackle the current ticket and then pick up the next one. But with this approach, it is easy to lose track of the bigger picture.
That is why I think it is crucial to plan what you would like to learn and in which direction you would like to grow, for example, within a year. A yearly development talk is definitely a good moment to address this topic. Additionally, it makes sense to catch up with your manager regularly on the progress and see if you are on track with your goals.
Pair-programming
While it is common to think that this approach might mostly benefit more junior data scientists, I do believe that it can be a great learning experience for both sides. The more junior person can accelerate their development by learning from the experience of their more senior colleagues.
On the other hand, I find that the more senior colleague can benefit a lot as well. For example, having to explain something might be out of their comfort zone or they might realize that they are not 100% sure about something in the first place. Or maybe they will hear from a more junior colleague about a model or tool that was released recently and they were not aware of it. Based on my experience on both sides, I think it is a win-win.
Mentoring
Similar to the previous point, I think mentoring can be a great opportunity for both the mentee and the mentor.
By having a mentor, you can learn much faster, as they will support you in your learning journey and might point you in the right direction in terms of your career progression. They will most likely have already faced similar challenges to the ones that you are facing, so such conversations can be invaluable.
For data scientists just starting their journeys, I would advise actively seeking out a mentor. In some companies, there are already processes in place for that, and you might even get a mentor assigned to you when you join. If that is not the case, I would strongly suggest finding one yourself. I'm quite sure most people would love the opportunity to help out if you ask them!
Additionally, I believe that mentors can also learn a lot. By trying to help their more junior colleagues, they learn the importance of listening, understanding, and providing helpful suggestions. It is an opportunity to show their empathy, as they should be able to easily imagine themselves in the very same situation. Not to mention that for most people it is a very rewarding experience.
Brainstorming sessions
Before kicking off a project and diving straight into implementing the solution you came up with, it might be a good idea to have a brainstorming session with your colleagues. Pitching ideas and discussing potential approaches is always fun, and you can learn a lot just from explaining your idea and why you think it's the correct one for a particular problem.
Sometimes, you might be the only data scientist on your team. If that's the case, I would advise you to look around your company and see if there are other data folks in a similar situation (alone in the team or in a very small team). Then, you can all join forces for such brainstorming sessions and help each other out while sharing knowledge.
Data science community
I realize that this may only be possible in larger companies that have multiple Data Science teams. However, even in smaller companies, it should be possible to gather all data professionals (data scientists, data engineers, MLEs, data analysts, etc.) and create some kind of community.
I believe a data science community can be a great asset to the company and all the individuals taking part in it. It can host events such as knowledge exchanges, training sessions, meet-ups, or even casual gatherings, during which data professionals from different teams simply sit together somewhere in the office and work on their own tasks while having the opportunity to meet others and learn about their work.
Another possibility is to have a DS community newsletter or a Slack channel where people can share interesting updates on what they are working on, perhaps share new models/tools they have discovered, or provide updates about upcoming conferences. Such communities also offer the option to seek help or opinions, as someone in your organization may have faced similar challenges to the one you're struggling with, and maybe there is no need to reinvent the wheel. There are definitely tons of possibilities!
I consider actively participating in such a community, helping out with initiatives, or even suggesting new ones, to be a great growth opportunity for a data professional at any level of experience.
Knowledge exchanges
I have already mentioned knowledge exchanges when discussing the DS community. I believe KX sessions are invaluable for data scientists for several reasons:
- During a KX session within your team/domain, you can dive deep into a topic and discuss the nitty-gritty details of the problem someone was working on. I think it's great to cover the solution that worked in the end, but also the ideas that had lots of potential but did not make it, as those often provide as much learning as the approaches that succeeded.
- During a KX session within the company/community, you can learn what other teams are working on. For example, you might be working on time series problems in your team. During such a session, you can learn how your colleagues are approaching problems from other fields of data science, such as recommendations, search, fraud detection, etc. While you might not be familiar with all the intricacies of these fields, I think it is extremely interesting to learn about them and also see how they are helping with the challenges that your organization is facing. And who knows, maybe they will inspire you to join forces with the other team and learn more about some of these areas!
- If possible, it would also be great to organize knowledge exchanges with other companies. While that might not often be possible due to a conflict of interest (e.g., two banks might not want to share their latest approach to fraud detection), there are always possibilities to explore. For example, you might not be the only company using time series forecasting for your demand planning. While there is no clear conflict of interest (e.g., one is a grocery delivery platform and the other a clothing store), you can definitely compare notes and learn from each other. For me, it has always been extremely interesting to see how other companies approach problems, what their tech stack is, etc.
You can learn a lot from all of the above both as part of the audience in such sessions and also as a presenter. I highly recommend presenting at such sessions, as you can not only improve your presentation skills but also hear very valuable feedback or ideas that you might not have considered.
Give and ask for feedback
I can't stress enough how important feedback is, both giving and receiving. Most likely, you will receive quite a lot of feedback during your yearly development talks with your manager. But you can definitely do better than that! Asking your colleagues/manager for feedback comes with many benefits:
- You receive feedback more frequently, so you can learn faster and adjust quickly if something is not working out as it should.
- You show that you are eager to learn and continue growing not only as a professional, but also as a person.
- Some useful feedback loses its meaning when delivered with a delay. As you can imagine, a lot can happen within a year. People might not remember the feedback they had for you regarding a project you worked on a few months ago. Or perhaps some people you worked closely with have already left the company, and you may not hear what they had to say.
Having said that, also try to give feedback to others and do not wait for fixed moments in the year when it is customary to give it. While some people might be shy about asking for it directly, I am sure they will appreciate it once you give it to them, and they might start sharing their own feedback with you as a result. Again, it's a win-win.
Personal development time
From what I've observed in the industry, many companies offer Personal Development time as a job perk. Under different names, it essentially means that you have some time during your work hours that you can spend on learning new things. While it may come in different formats, such as every Friday afternoon, a full day every two weeks, or once a month, the idea remains the same – you can use that time however you see fit to develop skills that are not necessarily connected to your daily projects. For example, as a data scientist focusing on NLP, you might want to explore classification problems simply because you find them interesting and you'd like to learn more.
Something that my team recently explored was an "out-of-the-box" week, which is a somewhat extreme example of personal development time. Together as a team, we spent an entire week (cancelled all meetings, and blocked the calendar) working on solving a specific problem in a hackathon-like setting. While some team members were already somewhat familiar with the problem, for many of us, it was entirely new, and we could come up with new ideas benefiting from a fresh perspective.
While convincing some people in the company that X data scientists should spend an entire week working on a long-shot idea might be a tough sell, I believe there are many benefits. These include team building, knowledge sharing, learning together, and ultimately generating new and valuable insights. So, while at the end of the week, we didn't necessarily have a new model in production, many of the insights we discovered were followed up on by the people working on the problem. Eventually, after some refinements and iterations, a solution we began exploring during the hackathon week ended up in production.
Working with other teams
I already hinted at this while describing the benefits of knowledge exchanges. I believe that temporarily joining another team for a joint project is a great way to learn new things. For example, as part of your personal development time, you could take on a small project with another team to experience hands-on what they are working on and how they are approaching solving data science problems. You might acquire useful skills that are not part of your daily job. And you can bring some of that knowledge back to your team!
The details of how to pull it off depend on many circumstances, such as your team size, your current workload, time-sensitive projects, etc. Ideally, you should join for a day or two a week to get the most out of this experience. However, even if it's just one afternoon a week, don't get discouraged and try to find such opportunities, as they will certainly pay off!
Conferences and trainings
Most companies offer a training budget of some sort. I'd definitely recommend using it to explore the possibilities and either improve your current skill set or learn something new.
Starting with conferences, they offer a great opportunity to learn about what other companies are doing, what difficulties they are facing, and how they are attempting to solve them. On top of that, attending conferences offers great networking opportunities where you can meet like-minded people working on problems similar to yours.
As for trainings, some companies offer in-house trainings led by either some of your colleagues or third parties. During such trainings, you might learn quite a lot about a particular topic, for example, data visualization. These are definitely super useful, especially at the beginning of your career. You might also be able to attend personalized training, which will focus on a skill that you'd like to develop, for example, putting ML models to production, writing production-ready code, or maybe leadership skills.
Don't neglect the non-DS skills
I can't emphasize enough that you can learn a lot of non-DS skills while working on your regular projects. Such skills might not be as easily visible to you, as they are not as clear-cut as learning a programming language or mastering a new API. Some examples of non-DS skills include:
- Stakeholder management
- Translating business requirements into data science products
- Identifying opportunities to leverage data science in your organization
- Working as part of a team
- Explaining complex technical concepts to non-technical stakeholders
- Project management
- Giving presentations
If you have attended some job interviews, you will quickly realize that questions about these skills pop up as often (if not more frequently) than some hardcore technical questions. That's because the answers to most of the technical questions are a Google search away, while it takes time to master the skills I have mentioned above. So my advice would be to start as soon as you can!
Wrapping up
I hope that after reading this article, you have picked at least one thing which you can use to keep on growing as a data scientist!
And how do you stay up to date with the latest developments in the field and how do you keep on learning? I'd love to hear your ideas, so please let me know in the comments or reach out on LinkedIn or Twitter!
You might also be interested in one of the following:
Experiment Tracking & Hyperparameter Tuning: Organize Your Trials with DVC
Use Rust's Speed to Install Python Libraries Up to 100 Times Faster
All images, unless noted otherwise, are by the author.
