Nothing Is Beneath You: Embracing Every Challenge as a Data Scientist
“This is beneath me.”
“This task is demeaning.”
“I’m overqualified for this.”
“This task feels insulting.”
“I shouldn’t have to do this.”
“This is not in my job description.”
“This work is too trivial for me.”
“This isn’t a good use of my abilities.”
I’ve heard these words way too many times, mostly from individuals perched precariously atop the Dunning-Kruger mountain. Surprisingly, these words sometimes come from seasoned and battle-tested engineers or scientists, who should know better.
Far too often, my graduate students express similar thoughts. They question why we start with a simple linear regression in closed form when they could just use a Transformer to forecast product demand. They wonder why they must keep plotting dataset feature distributions and “massaging” the data, even when we have powerful representational learning techniques or foundation models at our disposal.
In industry, I’ve repeatedly encountered the eagerness of some to label themselves “Data Scientists” after grasping a few machine learning techniques and spending a handful of hours training deep learning models in a couple of business scenarios. The reality is that mastering a few algorithms doesn’t entitle one to such a title.
I firmly believe that no task is beneath us. If testing our deep vision-augmented safety systems requires me to pose as a prone person in distress on a client’s bank agency floor, I’ll do it — even if some might say that’s a tester’s or systems technician’s job, not a Chief Research Officer’s. If you need to write a production-ready microservice, complete with a Dockerfile and a Kubernetes deployment manifest, you must do it — even if that’s typically a Machine Learning Ops Engineer’s role, not an Executive Data Scientist’s.
The truth is, there’s no such thing as work that’s beneath us or a waste of our skills. Yes, efficiency is crucial, and we mustn’t confuse inefficient resource management with the necessity of a hands-on approach and continuous learning. This holds especially true in fields like computer science and, undoubtedly, data science.
Let’s “backprop” a bit: To be a good (and proven) Data Scientist, you need to contribute to real-world product creation. This requires understanding beyond just linear algebra, programming, and statistics; you must also hone your skills in business analysis, systems architecture, DevOps, and product lifecycle management. Gaining hands-on experience in all these areas is essential, even if it means getting out of your comfort zone and getting your hands dirty.
Continuous experimentation and experience accumulation are vital as technology evolves. Nine years ago, my course curriculum didn’t include dev-containers. Four years ago, I hadn’t yet integrated LLMs. Six years ago, we used word embeddings, few-shot intent classifiers, and precompiled text corpora; now, we leverage foundation models. The world changes, and technology changes — often at an accelerated pace.
In Liu Cixin’s “Three-Body Problem” series, the concept of accelerated human evolution is explored from the perspective of alien civilizations observing humanity. The idea is that human technological and societal advancements follow an exponential, not linear, progression. This rapid acceleration makes humans particularly dangerous and unpredictable to the Trisolarans, the alien species central to the series.
As data scientists, I do believe we must embrace this rapid evolution and continually expand our skill set. No task should be considered too trivial or beneath us; every experience is a growth opportunity.