How to stand out as a great data scientist in 2021

5 Tips to help you stand out from the pack and get you hired from an experienced data science leader

Introduction

Data science is a highly attractive career for many reasons — so the competition can be tough. Some of the tips below that make a great candidate stand out might surprise you!

I’ve written this article for a variety of audiences:

• You might just be starting out and trying to land your first role
• You could be in a tangential technical field and want to make the switch
• Or you might already be an experienced data scientist looking to brush up on your skillset.

Why is it so hard?

This is a question I often get at user groups and community events for aspiring or early career data professionals. It’s a competitive landscape and the barriers to entry are changing — it’s no longer a requirement to need a PhD like it was when I started. As competition increases, flat-out technical ability becomes harder to differentiate yourself on.

Why listen to me?

I’ve been there — on both sides of the table! I’m an experienced data scientist that’s taken the steps through senior and team leader roles, all the way up to director level positions in tech startups.

I progressed quickly through my career and believe the points outlined below contributed significantly to that.

In my more senior roles, I’ve interviewed and hired dozens of candidates — from the talented and quiet professional to the weird and wonderful (ask me about the skateboarding vampire some time). I’ve interviewed for a wide range of roles including:

• Data scientists
• Machine learning engineers
• Data engineers
• Analysts
• BI developers
• DevOps engineers, CRM consultants, project managers…

In all that experience, there are key themes and skills that typically stand out (SPOILER: they’re mostly non-technical skills). I’ll share my take on them below and try to give some advice on how to improve them.

Technical skills

Now to get started, I’m not going to go through what algorithms you should know, what statistics books you should work through, or which machine learning stack you should choose. There’s a load of courses out there, many of them free, that’ll give you a great grounding in the field. I learnt during my PhD through free sources like this but there are many ways to get the right experience.

Broad or deep?

Is it better the be a generalist or a specialist? This isn’t a question specific to data science, but applicable to many roles tech. There’s also no definitive answer to this, it really depends on the kind of organisation you want to work in. If it’s your dream to land a role in a renowned technical team or research group with a focus on one thing — learning every flavour of machine learning model won’t work as well as really deeply exploring that niche. The field is too big to be an expert in everything (even keeping up with the research in one part can be difficult enough).

If, however, you want to be part of an organisation's first steps into data science or you want to work at a fledgeling startup, then having a broad perspective is usually best. Knowing the tools and techniques across the industry is going to allow you to solve a wider array of problems and more readily know how to unlock value from data when needs and requirements change.

Great candidates often have T-shaped experience — a broad knowledge and familiarity across the domain and a specialism in which they really shine (image by author).

Stand-out candidates are often T-shaped. They have a good understanding of the broader domain but have a specialism in one thing. This is often a more rewarding approach to your career too! It allows you to focus on the one thing you enjoy and really highlights your strengths, while still being able to jump into other types of projects.

Tip #1 — Being T-shaped in your expertise will allow you to concentrate and showcase your specific expertise but still contribute to a dynamic team dealing with a wide range of tasks.

Show your weaknesses

People often think you have to know everything and show it — especially when it comes to interviewing. This really isn’t the case and those that try it often interview very badly. I feel a real sense of relief when a candidate outright tells me they don’t know something — it builds trust and rapport.

Data science is a big and complex field. No one is an expert in it all. If statistics aren’t your thing, be open.

Like many, you might be a self-taught programmer, in which case don’t stumble your way through the coding tests regurgitating exam answers you crammed the night before without really understanding why.

Just getting a foot in the door by faking it at interview will lead to an uncomfortable first few months should you get the job.

Tip #2 — Be honest about your weaknesses and the things you don’t know. It’s often very easy to tell when someone is bluffing and you really don’t want to end up in a role that doesn’t allow you to shine.

Software engineering

The vast majority of people that apply to a data science role have spent all of their time learning machine learning, statistics, coding, and maybe some visualisation skills. When you start to prod them about design patterns or software development approaches they come completely unstuck.

There’s a big difference between coding hacky little models in Jupyter notebooks and building robust machine learning workflows that can be easily packaged, tested, and scaled into production.

Take some time to learn the fundamentals of software engineering. Even if you won’t need to use it all that often, you’ll interface with software engineers throughout your career. They’ve devised solutions for many of the problems early career data scientists bump into so learn from them. Furthermore, if you’re hired into a technical function within a company, there’s a high chance your department head or CTO was a software engineer originally — being able to talk their language will do you good.

Getting started can be daunting, you can’t go wrong with starting to learn the major themes across the following two books though:

Tip #3 — Learn the fundamentals of software engineering. It’ll make your work easier, your code more robust, and allow you to better relate to other parts of your organisation.

Up and running

For the majority of organisations, the most comprehensive and advanced models are worthless if they can only be run by data scientists. A lot of data science teams and machine learning projects fail because they can’t get beyond the exploration stage.

You don’t need to be an expert in containers and orchestration (unless you want to be an ML Engineer) but turning your models into value is a key part of the role. I almost always ask the question:

How will you get your model into the hands of a non-technical user?

In one of the best interviews, I ever held the candidate pointed me to some code they’d prepared and a very simple web app that hosted the model. I was immediately able to play with it and frame questions around what they’d done.

If you’re looking at getting started with MLOps I wrote a short series walking you through an end-to-end example here:

Tip #4 — Develop an understanding of the methods and tools to deploy your work into a production environment. Knowing basic MLOps will indicate that you understand model development is only a small part of the job.

Prove it

Sometimes, even getting to the interview can be difficult. This ties in well with the example given in the last tip. There’s nothing more powerful on your application than evidence that you have done this kind of work already.

If I’m looking through CVs and stumble across one with a GitHub link I’ll always check it out and often dedicate more time than I’ve allotted to that individual. It doesn’t even need to be wholly original projects, seeing how you tackle problems is more important. If you’re just starting out and don’t have a large portfolio of work, put your coursework up and write a commentary about how you worked through it. This will help you guide the interview before it’s started, as you're interviewer will almost certainly ask about it.

There are plenty of other routes to show your work too. Consider blogging as you learn (Medium is super easy to get started with). Don’t be afraid to get in front of people online through YouTube or at meetups and events — this will help brush up on some vital soft skills too.

You don’t have to be an expert to contribute to people’s learning. If you’re unsure on this or don’t know where to start, I highly recommend reading some of Austin Kleon’s work like this article:

Or this great book goes into it in more detail:

Tip #5 — showcase what you know. Having projects on GitHub, a blog talking through some of the things you’ve learned, talks and events or an active YouTube channel will make you stand out above the crowd.

Conclusion

Today I’ve tried to share what technical skills, for me at least, make a great data science candidate stand out. If all of this is new or daunting to you, pick one and start small. If you start blogging and adding to GitHub as you go, you’ll begin to sharpen your skill set and gain confidence.

It can be difficult to get the right role for you. Many people will appreciate a more rounded candidate that can interface across the organisation. Simply knowing the most algorithms or latest tools isn’t the best way to differentiate yourself in getting hired.

Hopefully, these tips will get you hired. If you have any feedback I’d greatly appreciate hearing from you. If you’re also a hiring manager in this space I’d love to know your thoughts.