3 Lessons I Learned a Little Late in My Data Science Career
Aspiring data scientists often ask me how I became a Senior Data Scientist while so young and how they can do the same.
The most reliable way to accelerate your career is to learn from those ahead of you. And then apply those learnings in your career.
I’m constantly looking to learn something new every day. I’m not ashamed to admit that I don’t know a lot of things already. Instead, I’m curious to borrow the brains of those experts through their experiences, learnings, and mistakes.
Certain lessons are better learned early in everyone’s career, but things don’t always go as planned.
In this article, I’ll present three crucial lessons I learned a little late in my career. Still, when applied your earliest, it has the potential to transform and accelerate your data science career trajectory.
1. Domain Knowledge Is as Important As Our Technical Skills
When we start learning data science, we mostly focus on technical skills such as exploratory data analysis, data cleaning, machine learning algorithms, statistics, linear algebra, and programming. Yes, this was a comprehensive list, but we’re missing out on something more substantial.
If we look at the data science Venn diagram, domain expertise is a crucial component of it. Yet, we ignore it as it doesn’t exist.
The reason behind our ignorance is that we rarely get access to domain expertise when we are in the learning stage. How do we get expertise in retail, financial, or anything unless we start working at a job?
I didn’t have any domain expertise until I started working in my first job.
In my first job, I learned about the use of AI in agriculture and animal husbandry. My boss even inquired if I was okay to work on data involving beef (the cow is viewed as a sacred animal, so even meat-eating Hindus may not eat beef.)
And when I started working, I realized there was a lot to learn from the domain-specific research scientists that were useful in model building.
Since then, I’ve worked on several domains such as smart cities, aquaculture, retail. It won’t be an understatement to say that every project had a massive domain learning curve.
We can never acquire this while learning and building models on the Titanic or Iris dataset. Some things definitely need to change.
How to do this early:
While in the initial stages of our learning journey, we might not have the luxury of working in a company. Hence, we have to find unusual ways to acquire these skills through the internet. Some steps I personally have followed:
- Pick any domain you are interested in. (e.g., Healthcare, Finance, Retail, etc.)
- Instead of common projects, focus on building a project similar to a real-world project in that particular domain. ( e.g., Using a diabetes dataset to build a project imitating a real-world scenario.)
- Watch YouTube videos and read scientific papers to understand the domain-specific terms and practices followed in the industry.
In an aquaculture project I worked on, I didn’t have access to domain experts in the team. I had no option but to research by watching YouTube videos (like this) and surfing the internet on the practices followed in fish farming.
I probably have gained only the minimal required knowledge to deliver the project, but that’s fine. Strive to find the right balance with the resources at your disposal.
2. Start As a Generalist to Eventually Become A Specialist
When I started learning data science, all I saw was most data science experts were specialized in one or two fields like NLP, computer vision, time series, MLOps. There have been reports that data science generalists won’t survive in the industry for long.
So as a beginner, I decided that I’d need to specialize in something despite having many unanswered questions.
- Do I know what I want to specialize in?
- Do I know what my strength is as a beginner in data science?
- Do I know if there’s sufficient demand for the niche I pick in the long run?
I picked a niche for the sake of specializing, and not so surprisingly; I got nowhere. Then something dawned on me.
Did these experts become specialists straight away as they broke into data science?
There was no need to be a specialist as a beginner. In contrast, I needed to focus on becoming a generalist first. I had to understand the entire pipeline and try different domains before I settle on one. So I did just that.
I worked on translating business problems into machine learning solution blueprints, developing dashboards, presenting insights, developing models, creating MLOps pipelines, and more. Basically, anything that came my way.
Fast forward to now, 3 years in, I know what I’m good at, and I’m focusing on specializing in it for the next 2–3 years.
How to apply this:
- Early on in your career, please keep an open mind to opportunities.
- Start by working on everything possible in the end-to-end data science pipeline.
- With time (read, 2–3 years), you will have an intuitive understanding of your strength and interest.
- Pick specific niches within data science strictly based on your strengths (e.g., NLP for social media, Computer Vision for Retail, Production-ready MLOps, etc.)
- Remember, the field is constantly evolving; continue developing advanced skills into your chosen niche.
When you follow the “generalist to specialist” approach, the benefits are two-fold. First, you understand the entire lifecycle, which is essential when managing junior data scientists in your team. Second, you have built specialized skills in a domain that will help you survive in the industry for the long run.
3. We Learn Our Best When We’re Expected to Teach
This is my secret of learning better constantly. Feel free to steal it if you want.
In my final year of undergraduate studies, I started learning data science on my own. I’d go through Andrew Ng’s lectures on Coursera and build projects. I was learning from the best, but I had no guidance on going in the right direction.
Chances are you feel the same. You know these are the popular courses most recommend, but you’re not sure if this whole thing will pay off. You wish you had a support system to learn and hustle together to make it big in data science.
For the whole first year of this journey, I felt the same until I was asked to deliver a workshop on Structured Deep Learning. Suddenly, my learning had a purpose: a bunch of colleagues will learn from me. In an instant, I recognized the difference in my learning process — I can deliver the same workshop even today. I remember it all.
Fast forward to now, I have delivered many more workshops, written tutorials about simple and advanced topics.
Now I have you guys who expect me to teach whenever I learn something new. That gives me a whole new level of purpose to learn better.
Thank you, yes, you, for reading, interacting, and being my purpose behind this journey.
How to do this early:
When you’re progressing through your learning journey, indeed, you don’t have much to teach. We all know that, so here’s what you do:
- Start by sharing your experiences on what you learned, what worked with your friends, or on platforms such as LinkedIn, YouTube, Medium.
- These platforms help you find enthusiasts at a similar level who have the same problems as you. A small group of peers who have the same goal as yours will become your support system.
- Teach your friends on any simple topic (say, linear regression.)
- Slowly as you gain expertise, try delivering a presentation to a small group. If you're still studying, it could be for your class; if you’re working, it could be for your colleagues who are interested in data science.
- You know these opportunities will always pop up — you need to be ready to grab them.
- Obtain feedback and improve. Unlike me, when you start early, you’ll have a 10x better learning journey.
Sharing knowledge always solidifies your understanding of the same. Teaching is the best motivator to learn your best, period.
Thank you for reading so far. I genuinely hope this was of use to you, especially if you’re starting to learn data science.
I extensively write about my learnings and experiences in the world of data science on this platform and LinkedIn. The community was helpful when I started, and this is my way of giving back.
The whole point was to help you accelerate your data science career. In summary:
- Acquiring domain knowledge is as important as growing data science technical skills.
- In data science, start as a generalist to eventually become a specialist.
- Learn data science to teach someone. You and the community will benefit tremendously.
Breaking into data science can be challenging; nobody said it was easy. But many are doing it every year. There’s nothing special about them, nor me. All of us learned from the experts and applied it on our journey.
Now, what’s stopping you?
For more helpful insights on breaking into data science, exciting collaborations, and mentorships, consider joining my private list of email friends.