Hacker News Reader

Comment by wallscratch | Hacker News Reader

As a current PhD student in computational neuroscience, I think most of the academics I interact with tend to drastically overestimate the value of their “data science” skills to real jobs, and underestimate the value of software skills beyond prototyping.

View on HN

I spent my Ph.D. coding in MATLAB, as my Ph.D. was experimental and MATLAB was mainly used for data analysis and plotting. On the other hand, I had a full conda package with complete documentation up and running by the end of my postdoc. Though this was a side project, the time spent on this was incredibly valuable. I did not have to prove to anyone I was proficient in Python; my GitHub was enough. Additionally, the vast majority of researchers' extent of Python expertise was limited to disjointed Jupyter notebooks - while I had a running package with extensive documentation.

I got 3 job offers just based on my package itself - while very few non-academic jobs were interested in my publications. The fact that I had a few first-authors in reputed journals was enough; nobody was interested in their contents.

At my workplace we assume if you're a good Matlab programmer, you can teach yourself Python in a jiffy. What's a bit tricky is that every resume mentions Matlab and Python. Having a public repo is a useful way to show what you can do.

This is the key. Everyone claims Python/MATLAB proficiency, yet there is honestly a vast competency spectrum. A polished public repo, in my experience, will really help you stand out.

Industry is still moving from excel / spreadsheet modeling to tableau and python and alteryx in the workplace. Knowing python today is knowing excel 20 years ago.

Most companies aren’t ready to do real data science, and the transition is as much an organizational problem as it is a technical one.

What makes difference is the autonomy in the job. Are you hired to work as a cog, or do you have power to influence what you do? In the latter case, data science skills multiply the amount of impact your work has (assuming you have ideas).

It's the difference between, "I explored these 5 scenarios last night, none of them is pans out" vs. "This idea seems interesting. We need three weeks, and small team to explore this idea that might have potential."

My impression is that most companies don't know how to appreciate and take advantage of these data science skills, so it's not so much that the skills aren't as valuable as academics think, it's more that they're not utilized in an effective manner in real jobs.

Reply on news.ycombinator.com