Image by Author
Every day, I wake up at 8 am, open my laptop, and let AI do my job while I browse Twitter. At least, that’s what some people think a data science job looks like these days.
And while I wish that were true, my job is slightly more complex than copying and pasting AI-generated code.
I work as a product data scientist at a large tech company.
What I Do As a Data Scientist
Here’s what I do on a daily basis:
- Work with product teams and run A/B tests to decide whether a feature should be launched.
- Design success metrics to analyze the performance of products that we release.
- Build ETL pipelines and dashboards to monitor our features.
- Help the business iteratively improve our product and user experience.
Note that a lot of what I’ve described above requires an intricate understanding of the business, along with strong collaboration with product, design, and engineering teams.
The job is a lot less predictable than people expect. In fact, I can spend an entire workweek designing a single success metric to capture how well our product is performing. This includes:
- Understanding what exactly we need to measure and why it matters to the business.
- Working with engineering teams to understand how they can track the metric we’d like to capture.
- Coming up with logic to calculate our desired metric and turn it into an SQL query.
- Building a dashboard to showcase the performance of the metric.
If you aren’t a data scientist, you’re probably a bit lost at this point. And I get it.
It’s hard to explain exactly what we do to someone who doesn’t work in the field. For simplicity, you can think of a data science job as:
- 50% logic and math
- 30% product and business knowledge
- 10% presenting findings to stakeholders
- 10% coding
I might be oversimplifying, but you get the gist.
AI can help with the coding part (10% of the job) — which only comes AFTER the data scientist does all the product brainstorming, metric calculations, and business understanding.
AI Doesn’t Worry Me… For Now
AI still doesn’t have the reasoning capabilities required to do 90% of my job. And if I can leverage it to help me become a better programmer or even automate some parts of my workflow, that just makes me a more efficient data scientist.
It helps me get things done faster than I used to before.
If You Want to Become a Data Scientist, Do This
Data science roles are much more complex than the Internet makes it out to be. If you want a high-paying role like a product data scientist position, there is a lot of ambiguity that comes with the job. You need to figure out solutions to problems that the business has never faced before, and you must do this with the data you have. If the data points you need aren’t being captured, you’ve got to generate complex logic and figure out a different way to calculate your desired metric.
There is also a ton of ETL work involved, and I’d recommend taking a data engineering course or two before you interview for a data science position.
Takeaways
There still is a market for data science positions, particularly in the product realm.
The most useful data scientists (the ones who are paid the most) will be those who have an iron-clad understanding of how the business works, along with product, engineering, and domain expertise.
The closer you are to the business, the harder it gets to be replaced by AI.
Here’s what this means:
- If the company relies on you to make critical business decisions
- If you work closely with business teams and have a direct impact on the company’s bottom line,
You don’t have to worry about getting replaced by AI anytime soon.
 
 
Natassha Selvaraj is a self-taught data scientist with a passion for writing. Natassha writes on everything data science-related, a true master of all data topics. You can connect with her on LinkedIn or check out her YouTube channel.