Five hard truths about building a career in Data Science
Why do newcomers find it overwhelming to land their first job in Data Science
A simple google search for building a career in data science will throw a long and exhaustive list of skills in your face. From programming in python to applied statistics with proficient communication skills and making reports and dashboards. To add to this chaos, there is a large spectrum of jobs with different shades of requirements. For example, a fintech company may look for a Ph.D. in Mathematics whereas a product-based company would be in need of a shrewd Statistician to design and run their A/B testing experiments to drive decision science.
As a result, aspirants who are still not certain whether they want this or not, get depressed by this huge heap of required skills.
Truth #1 - Nobody can do everything. Become data literate and then narrow down.
The truth is there is not a single Data Scientist or Analyst who possesses all these skills. A good bunch of them specialize in 2-3 skills on top of their basic data literacy that make them irreplaceable.
A good approach here would be to acquire enough mathematical and statistical knowledge to do the following:
Expanding the available methods to apply. If you’d not know something exists, how will you apply it?
Choose the right methods - out of the available techniques, assessing which would work well in your scenario.
How to apply those methods - understanding the complexity of applying a technique to your problem. Be able to configure and optimize.
Truth #2 - The paradox of experience
Another challenge with getting a job in this role is that most companies want experienced Scientists who are can drive decision-making and refine their business models. Then how do newbies break into DS?
You must have come across the memes on the paradox of needing the experience to get experience. Those memes hold true for DS. The only solution is building a solid work portfolio that showcases your expertise in that domain.
Truth #3 - Surge in newly graduated data scientists
We have graduates coming in from degree programs in Universities, boot camps, or MOOCs. The number of available eligible candidates per job has shot up from 25 resumes per job to 100 resumes per job and it is only going to grow with time.
This isn’t something that you should worry about but it is something to be wary of. It should push you to have a smarter plan.
Truth #4 - Programming is inevitable to become a good Data Scientist
Up to a certain extent, you can use Excel or other BI tools(like Tableau) for your analysis or even modeling but they don’t offer the flexibility to run multiple analyses with different datasets.
Reproducibility is another major concern which is why we are now looking at teams(check out iterative.ai) setting up version control for data and ML pipelines. No one tool offers this as of now. Even if it did, the number of possibilities is immense for any one tool to handle.
Truth #5 - No matter where you come from, domain knowledge is key
The last company I worked at, sends research papers to its Data Science candidates to check if they can get their head around biomedical problems and data. This exercise informs the team about the ability of the candidate to make decisions and drive growth. Business understanding is basically the intersection of data science paradigms and practicalities of real-world problems.
It helps you define the right questions, prioritize the right product line, choose the right data sources, measure the right metrics and in turn help the business grow faster.
A big chunk of newcomers finds it hard to develop domain knowledge in conjunction with experience in the field.
That’s it for this week but hang tight as I demystify each of these challenges every week.
Announcement
In the last week of April, I will be doing a free webinar on Landing a job in data science where I will be sharing tips on how to strategize, write resumes & cover letters, and jobs you can apply for in different parts of the world.
To help You better, I’d need you to fill out this form. It’ll only take 1 minute!
Great article, Harshit! Very insightful! :)