I worked as a data scientist and that was the worst job I have ever had.

I thought that a data scientist position would be a dream job. I believed in “the sexiest job of the 21st century” hype. Silly me.

Because of the hype, I spent a long time learning data visualization and machine learning. I started blogging about data science in June 2018. I got the data scientist job one year later. And… I quit it after two months.

Between June 2018 and the day I left, I wrote over 100 articles about machine learning. I had learned a lot, but data science wasn’t for me. Why did it happen?

Ultimately, I didn’t like the job. I didn’t like anything about the job - the commute, the office, and the job itself. The team was cool, though.

This happens when you have online interviews and don’t see the working conditions before your first day at work.

The neighborhood

Imagine an old artisan district in a large city in Poland.

It is a district full of old tenement buildings with shops on the ground level and apartments above them. Of course, not all of them survived World War II. To replace the ruined structures, the Polish People’s Republic rulers built Soviet-style housing blocks - gray monuments of sadness and despair. Nowadays, we paint them in bright colors in a futile attempt to make them look less like the human equivalent of battery cage poultry farms.

To spice up the mix, we fill the unused space with ridiculously overpriced, modern apartment buildings with walls so thin you can hear the neighbors talking or flushing the toilet.

You know what the area looks like, but what about the people?

In the 90s, the place I described was one of the worst districts in the city. It was infested with drug dealers, aspiring rap musicians, and juvenile criminals. Occasionally, all 3 in one person.

Today, it became the hipster area. You can easily find vegan restaurants, craft beer pubs, and dozens of places selling kombucha. If you are not looking for trouble, you should stay near those hipster venues. Wandering off is not advised. When you turn onto the wrong street, you will find a different place - the remains of the 90s.

A few meters away from the hipster spots, you can time travel to the 90s. You may step on broken vodka bottles scattered on the sidewalks or meet a drunken guy asking whether you have any problems (which means that you may have one in a second). If you are particularly (un)lucky, you may see a police car stopping abruptly in the middle of the street and witness officers sprinting into one of the buildings.

At the end of one such street stands a school. There is nothing extraordinary about the school besides one online review. The one-star review features a picture of a teenager having bruised face and a black eye. In the comment, he wrote: “They beat me.” (The photo was already removed, only the text remains.)

In front of the school, on the other side of the street, stands a vine-covered residential building partially repurposed as a coworking space. This is where I worked.

We were told never to leave our computers in the office. Are you surprised?

The workplace

The office space was an apartment with furniture removed and replaced with desks, office chairs, and cool-looking but quite dim lamps.

There are old buildings with large windows and bright rooms. Our coworking space wasn’t one of them. We were renting a tiny room with only one window.

Like many rooms in XIX century/early XX century buildings, the room was quite tall. As you may expect, it was cold in the winter and hot in the summer. Fortunately, I didn’t stay there long enough to experience both.

We had five small desks in our room with no space between them. We all had B2B contracts, so Occupational Health And Safety Regulations did not apply. We weren’t employees, so we “enjoyed” being very close to our colleagues. I’m glad we could work from home, and I was doing it quite frequently.

Naturally, our office could be worse. One day, I saw a Slack message sent by one of the founders who tried to disprove rumors about rats in the NYC office.

Our office was way better! As far as I know, we didn’t have rats roaming around.

The first week at work

During my first week at the new job, the company had an off-site event. We went kayaking. I love kayaking! I was lucky.

Except for one tiny detail.

It was a low-budget event, so we went to a cheap place. Which places are cheap? The ones whose business suffers because of dry weather. I spent half of the trip walking in the mud and shallow water, dragging the kayak behind me through the muddy riverbed, slippery rocks, and fallen tree branches.

This is fun! Seriously! I love it! But only if it happens once every 20-30 minutes. Not all the time for 4 hours.

Our adventure got even more exciting when a thunderstorm started in the middle of the trip.

So here I was. In the middle of the forest. Underneath the trees. Standing in the water. During a thunderstorm.

I was a scout as a 9-10-year-old child. If I recall my training correctly, the only thing I could do to make the situation worse was holding a five-meter-high metal pole in my hands.

We were lucky that day, and nobody got hurt. However, the other data scientist somehow managed to soak his mobile phone in an almost dried-out river.

Every following day must be better after such a first week at work. One may think…

The task

Our startup’s founders couldn’t claim they were “making the world a better place,” “disrupting,” or “revolutionizing” anything.

We were running websites with celebrity gossip, trivia, crime stories, etc., selling advertisement space.

Was this a viable business? Hell no. Almost nobody looks for such websites, so we had to buy tons of ads elsewhere to get any traffic. The idea was based on the old-school concept of buying low and selling high.

They needed data scientists for the “buying low” part. We were trying to automate bidding and get as cheap traffic as possible.

Why was I working there?

It was the only place where an inexperienced data scientist could get a job. We all already know why the experienced people didn’t want it.

Overall, was it a good business idea? Of course not. We were burning money fast. It was so bad we were celebrating every day when we managed to break even. It happened only once during my two months at this startup.

I know, I know. This is how startups work. It is true, but usually, startups pivot when their idea doesn’t work. At this place, we were doubling down on things that didn’t work in the past and expected to get different results this time.

Quickly, I realized I don’t enjoy training machine learning models nearly as much as I thought I would.

It was fun when I was learning it. Mostly because I was switching to a different problem every time I got a working solution. Not here. This was the real world, not a Kaggle assignment. I had one problem and one problem only: figure out the maximal amount we could spend on buying ads and still make a profit.

Worse still, I wasn’t as good a data scientist as I wish I had been.

I realized I enjoy deploying the model and building data pipelines more than training the models.

Soon, I focused on doing the data engineering part of the job. Nobody complained because the other data scientist didn’t know how to do it. Others perceived my behavior as stepping up to a challenge, not hiding in the comfort zone. Only I knew the truth.


At the same time, I read the “Career Superpowers” book by James A. Whittaker. In the book, he wrote the “Underachievement Manifesto,” saying:

This may seem harsh, but there is opportunity written all over it for those who aren’t too blind with ambition to see it. The ranks of any fields, no matter how mundane or exciting, are full of people who have stretched to get there. People not quite smart enough for medicine still practice medicine. People not quite dedicated enough to law are still lawyers. People who aren’t particularly mechanically minded still try to fix cars. This is why a good mechanic stands out, he or she is competing against people who aren’t really good enough to be there.

That was me. I stretched to become a data scientist, and I didn’t even enjoy the job. Soon, I switched to data engineering/MLOps and never looked back.

When I handed in my resignation, the team couldn’t understand why I quit a company where I could influence the product and do whatever I wanted (as long as it was related to buying cheaper ads).

Partially, this was the problem. Everyone was doing whatever they wanted (including the founders) instead of trying to earn money for the company. Apparently, having a startup or being employed in one was more important than making it a successful business.

I told them I didn’t believe they could survive the next six months. I was wrong. They went out of business after eight months.

I didn’t include the data scientist job in my CV or LinkedIn profile. I am pretty ashamed of it. Most people don’t notice the gap or assume I took a long vacation between the jobs.

If anybody asks, I tell them this story. So far, nobody has questioned my decision.

Older post

MLOps engineer, you will need those three books every day!

Don't reinvent the wheel as a MLOps engineer. The 3 books you must read in 2022

Newer post

Data engineers are data librarians or how to upgrade your data lake to 2500 BCE technology.

What can data engineers learn from (ancient) librarians?

Are you looking for an experienced AI consultant? Do you need assistance with your RAG or Agentic Workflow?
Schedule a call, send me a message on LinkedIn, or use the chat button in the right-bottom corner. Schedule a call or send me a message on LinkedIn