Five Tips for Overcoming Imposter Syndrome in the Data Science World

Hello there, friends! I thought I’d take a small break from the typical tutorial-oriented posts to cover a different topic that is still data science oriented and one that is underrated in my opinion. Of course, I’m talking about imposter syndrome. In case you are new to the concept, imposter syndrome is this idea that you feel like you aren’t good enough to be in a particular position. It’s not a concept localized only to the data science world, but in my opinion, I think data scientists and machine learning engineers get hit the hardest with imposter syndrome.

To share a quick bit about my personal background, I have been a machine learning engineer at a Fortune 50 company for about 18 months now. Prior to starting in this position, I had never been in any sort of technical role. My bachelor’s degree is in business administration, and my master’s degree is in organizational leadership. My prior roles included things like scrum master, business architect, and general business analyst.

It’s a long story of how I stepped into the world of machine learning, but my education path wholly revolved around alternative learning platforms. While I have no formal college degree in anything remotely close to computer science or machine learning, I learned my current skillset in the form of nanodegrees, MOOCs, and industry certifications. After building a portfolio of work and demonstrating my competence in the space, I was honestly surprised to receive a job offer to begin my role in January 2020.

But if I’m being totally honest… I felt completely out of place when I first began my role. Keep in mind that this wasn’t just my first job in machine learning; this was my first job in any computer science role. Period. For the first few months, I got very down on myself thinking that I would never be good enough for the role. I very frequently thought, “What am I even doing here?” There were many times when I seriously contemplated reverting back to a business-oriented role. It’s honestly only fear of embarrassment that stopped me from doing that.

I’m glad that I stuck with it though, because even though I would not consider myself an “expert” in pretty much anything, I’ve seen a lot of really neat success in my role. I am now team lead over a small group of other machine learning engineers, and I was very recently promoted myself to the “senior machine learning engineer” title. If you would have told me this back in February 2020 when the imposter syndrome hit the worst, I genuinely don’t think I would have believed it back then.

As somebody who genuinely wants to see others succeed in this field, I wanted to take some time to address this topic since behind closed doors, I know for a fact that I’m not the only one who has experienced this. Having mentored a broad range of folks in machine learning and data science, many of them have expressed this same fear. My hope with this post is to share some things to consider to help you overcome your own imposter syndrome. With that, let’s jump into five things to think about in regards to imposter syndrome.

1. Everybody starts new in any data science role due to domain knowledge and expertise.

Whether you write basic logistic regression models or fancy deep learning algorithms, the reality is that even if you understand the math under the hood, domain knowledge is absolutely key to success in any data science job. In that sense, everybody has to begin anew when they join a new job in a data science capacity. Even if a person simply pivots from one department to another within the same company, that person will likely have to learn a new set of domain knowledge, which obviously puts them at a disadvantage in the beginning. There is no escaping this even amongst the most seasoned of data science practitioners, so whether this is your first or tenth data science position, everybody has to start fresh with a new set of domain knowledge.

2. Many data science job postings are poorly written.

As a Lambda School mentor, I have sat down with some of my mentees to review some job positions they are interested in, and it’s no wonder my mentees get super daunted after looking at the open positions. There are two general issues I see very frequently with data science job postings. The first is that a job posting was clearly written by an HR recruiter who doesn’t have any experience working in a data science capacity, and the second is that the job posting will ask for an insane amount of skills. I have seen some postings that will ask that a candidate be proficient in Python, Java, Scala, R, C++, Kubernetes, AWS, and more. I work with a pretty smart group of folks, and let me say this: nobody — including myself — comes close to checking the box off on every single desired skill. And most employers don’t expect the full gamut anyway. Don’t let a laundry list of skills keep you from applying for a position you’re interested in.

3. Keep in mind how fast technology moves.

Technology, especially in the data science world, moves so fast that it’s highly likely that even seasoned practitioners will have to continuously hone their skills over time. It’s sort of wild to think that things like Docker and Kubernetes, which I work with on a daily basis, haven’t been around all that long. Even since I started in this role about 18 months ago, Amazon Web Services (AWS) has added a ton of new features to the SageMaker service, particularly in the form of SageMaker Studio. I personally have not had the opportunity yet to learn those things coming out of SageMaker Studio, but I know that if I don’t try to keep up, I will fall behind. The pace of evolution is not going to slow down any time soon, so remember that if you feel that you’re having a tough time keeping up, chances are that even the seasoned practitioners struggle to keep up, too.

4. Don’t be afraid to take care of your mental health.

This is a highly underrated thing to consider, especially if you are simultaneously coping with stressors outside of the workplace. For me personally, COVID-19 hit just three months into my role, right at the seeming height of my own imposter syndrome. While I wasn’t exactly afraid of dying from the virus, I had a lot of stress about the potential ripple effects of the virus. How would this effect the economy? What does this mean for my future employment status? Will this virus radically change the future of humanity? It was bad timing all around, and to be honest, I sought professional help. I feel like this something we as men in particular struggle to do because it feels like showing weakness. To be transparent, even typing these words here feels very vulnerable and uncomfortable, but I think it’s important to be honest with you if I can help influence you to get the help you need. I know how hard it can be to seek and accept professional help for your mental health, but it is a decision I do not regret at all and am very glad that I did.

5. Remember that all around expertise is an illusion.

I must say, I still experience a natural bit of imposter syndrome myself when I open a publication like Towards Data Science. I know how ironic that is given that I myself am a regular contributor to the publication myself, but I am constantly impressed by things people contribute to the data science community. The thing that I have to remind myself is that knowing that I myself am a contributor to the community but admittedly way under-educated in several aspects of data science, I have to imagine most people are the same way. That’s not at all being condescending to anybody. It’s just a reality that there is so much to be learned that I don’t think it’s even possible for a single person to learn everything. We all might have expertise in a narrow slice of the world, but I very highly doubt anybody is a master of everything.

That wraps up this post, folks. I hope you find it encouraging as you continue along your data science journey. Remember that we all have to start somewhere; nobody starts off as an expert. Thanks for checking out this post, and we’ll likely be back to cover more of the Terraform + SageMaker series in the next one.