So you landed a job as a data scientist.
You explore really interesting data, come up with a hypothesis, build a model around it, solve the problem and make millions for your company!
This is misleading. This is not the whole picture. This is a mirage.
Let me walk you through some mistakes I made in the first year of my job as a data scientist so that you don’t make those in yours.
1. Not knowing the “kind” of data science my teams needs.
Sometimes, you need to press pause on personal interests and think about what’s best for the product or team rather than just personal progress.
When I started off at work, I was seriously interested in NLP. Almost obsessed with it. I even picked a team that would allow me to work on something like language modelling for the existence of the product. I was ready with my prep on PyTorch, Tensorflow and other deep learning techniques like a Super Saiyan just ready to unleash a transformer model on the tech stack.
Little did I know that at that point in time the product didn’t require work on this part. There were other features that needed improving. The team was focused on other things. My goals weren’t aligned to the big picture. Me trying to train a model was miniscule compared to doing all of the other engineering work. Understand the big picture faster, make the team win. Use that success to fuel personal growth.
2. Not having a mentor.
Having someone who can predict what could change ahead of time can save you a lot of unnecessary effort and give you much needed cushioning when the going gets tough.
Looking back, I spent the first couple of weeks at work trying to become better, technically by figuring things out by myself. This isn’t all that bad, but having a mentor, who is not just the person assigned to be your mentor, can be a game changer. I’d say, the first month on the job, forget learning the tech stack, forget reading that paper, forget playing foosball just look for a mentor. Get those lunch meetings going from day one. Start by asking 10 people and follow up with them. It trickles down to 3 and finally 1 or 2.
It’s so easy to get caught up working on something that you forget to socialize and build bonds with people. The reason I stress that this is important for a data scientist is because it’s a growing field, the definition of the job itself changes a couple of times a year. Trust me, there are hacks to everything and having a mentor for a variety of areas within the company and outside of your company can give you much needed perspective and save you time on some avoidable errors.
3. Skewed knowledge.
Being an expert at something is great only when it’s coupled with the fact that you are willing to be good at other things as well. Be coachable.
This is the effect of something called the Icarus Paradox , which states that, “failure is brought about by the very elements that led to their initial success”. Deep knowledge about a topic can impress interviewers and land you a job, which you then sort of take to your team in a company. Another hard pill to swallow is realizing that becoming an expert at just one thing too soon actually puts you at a risk of not getting diverse projects. These are the things you learn if you have a mentor. Lol.
4. Being afraid to fail.
In industry, the right failure is considered a success if it saves someone else time. Because, time is money.
Okay by far, most useful mistake, really. There will be times when you need to take that extra step and try something new. Do it. Don’t wait for your manager or mentor to hand hold you. Be the trailblazer. If it fails and you make a total fool of yourself, there are two things that happen. Firstly, you discovered a mistake which someone else could have made so you saved someone else time. Document your process. Make it an experiment. Being a data scientist is all about trial and error. Second, you end up claiming a territory. If you experiment with say, graph neural networks and end up trying to predict something, despite failing, your initiative shows promise and that in itself shows that you’re willing to take risks and go the extra mile, which no one else did. The last part is really important and stands out to who ever is evaluating you.
5. Not learning everyday.
The biggest obstacle to learning is being intimidated by how much there is to learn. You’re not in university anymore, you don’t need to take 3 courses at a time. Start small.
There are >1000 people out there waiting to replace you. Let that sink in. If you have a job in this economic climate, you are privileged. A job as a data scientist? Extremely privileged.
Yes getting the job was tough, but dealing with not being on top of your game everyday on the job is tougher. It took me a while to realise that learning was not something I had to do in bulk. I no longer had to take a 5 credit course on Deep Learning. I just had to go through 1 article a day. Read 1 abstract a day. Brainstorm for 1 hour a day. Growing just 1% everyday, works wonders. Don’t be afraid of learning, it’s what you’ve been doing everyday since first grade. You’re programmed to evolve so embrace it. :)
I’d just like to say that, despite very openly confessing some of my mistakes during the initial months of my job, I did have a good time as a data scientist. I got the opportunity to work with some fantastic people, understand how large models go into production, scaling ML models, the business sense behind a technical decision and of a course lot of people management from my manager. I’ve learnt that admitting your mistakes can ensure someone else who is lost can possibly find their way and that’s what I was hoping to do with this article.
If you’re someone who likes to learn everyday and work on things that are changing more than a couple of times in a fiscal year, being a data scientist will keep you engaged and interested. It’s not easy and definitely not comfortable so, choose your battles wisely. Good luck on your job, young Padawan!
Feel free to reach out to me for anything and if like me you want a refreshing change of pace from your job check out the podcast I host. Cheers!