So you want to get a job in data science?
You’ve been doing all the courses and you’ve been making some pretty sweet visualisations but now what?
I was in Melbourne recently and had the privilege of being shown around one of Australia’s largest tech companies, REA Group.
Their main site, realestate.com.au gets 7 million unique visits per month — remember this is Australia, our population is 22 million. So 1 in 3 people visit the site every month. We love real estate.
From these 7 million unique visits per month, 300 million data points are collected, analysed and experimented with to provide a better experience.
When Nigel Dalton, Chief Inventor of REA Group, and I were discussing the timeline of REA Group, he said they’re moving away from a search space into a matching space.
Rather than searching for the perfect property, REA wants to use their data to help match you with the perfect property. In essence, using your past preferences to show you places you’re more likely to be interested in.
If you’re thinking this is something new, it’s not. It’s happening all over the web. Netflix’s recommendation engine and Facebook’s newsfeed do this. My brothers and I rarely search for YouTube videos anymore because their recommendations are so good.
Patterns. Patterns are everywhere. The 3-inches at the top of your skull is an expert pattern recogniser. It’s how a 20-year veteran real estate agent can look a home in their local suburb and guesstimate the value within a few decimals. But now our ability to replicate this pattern recognition with machines is catching up, in some instances, surpassing our grey matter.
There’s no way you or I could take in the 300 million data points and imagine bettering the experience for 7 million people. Maybe I could take in 10 and better it for 1 person and you 20. Then how do you match someone with a property based off of their previous searches? Data science.
What was I looking for?
When I was talking with REA’s most Senior Data Scientist I wanted to find out what it took to get on his team.
For clarity, REA has a data science team of 12 and data plays a part in almost every one of REA’s business decisions. There’s something beautiful about how much impact a relatively small team can have with access to the right tools.
“What do you look for when you’re looking to hire someone new?”
I’ve condensed our discussion into the following points.
Skills are a must. There’s no question here.
But what skills exactly?
Python or R is where you want to be.
How much Python? How much R?
There are no exact answers here. Being able to manipulate a dataframe is a great place to start. If you’re unsure what a dataframe is, don’t fret. Find out and learn about them, then keep reading.
You’ll have no trouble trying to find a data science or machine learning course online and a book or two to go with it. I learned this way and I’m going to continue learning this way. If you’re looking to get started, I recommend anything on Coursera or Udacity. Yes, they’re paid, but you get what you pay for.
There is no perfect set of skills or perfect course/book. Jump in and give it a try. Once you’ve learned a few things, start thinking about how they could be applied to something you’re interested in. It may take a while to get to this point, it may take a day. Either way, remember, learning a new skill is by definition hard. Expect it to be hard at some points.
“Most of our problems don’t require the same skills, the foundation is needed but not everything is the same.”
So you’ve been studying hard and now you’ve got a few skills. Great! Now it might be time to apply for some roles. But you read the job listing and are quickly disheartened by the requirements.
X years experience in R.
X years experience in data visualisations.
X years experience in PowerPoint (what?).
You’ve been learning Python. Damn. That role looked really good.
Wait! Don’t fear. Every job posting will have something like this on it. It’s the norm.
Whatever skills you learn, it’s unlikely they’ll ever fit the exact ones listed in the job posting. Apply anyway.
The beauty of these skills is their adaptability. And that’s what you’ll have to be prepared for. Problems won’t always come in the same neat little package.
This is where you can demonstrate what you’ve been up to on your learning journey.
Alright, you’ve been learning some skills and you’ve tried them out on some data. Everything didn’t go as planned.
Now you’re reading about a role you’d really like but again you’re thrown off by ’46 years experience with being a data ninja.’
That’s another way of saying, ‘we need you to be able to ride a bike on your own so we don’t have to teach you how to ride a bike with training wheels again.’
Well, guess what? You’ve already ditched your training wheels. You might be a little wobbly, but you’re still riding the bike.
But throwing analogies like I just did might not cut it in an interview or a cover letter. If they do, virtual high-fives to your employer (and you).
The internet has given us a menagerie of tools to show how creative we all are. A few came up in our discussion.
Ever wanted a corner of the internet you can make entirely your own? This is where that can happen.
When I first launched my blog, I was scared. After a couple of years its become a part of me, I love having my own space to share my ideas and anything else I’ve been working on.
You can code your own, but we’re also spoiled for choice with places like Medium, GitHub Pages, SquareSpace and WordPress. Write an article about the latest project you’ve been working on. Share an idea about how to find insights into an open source dataset.
If a post gets 0 views, don’t worry. It all adds to your proof of work. So if someone asks, “have you ever had experience with X?” you can say, “well not exactly, but I did write an article about how I dealt with Y and how I learned Z from it.”
You know what I’m talking about. The place where code lives.
If you’re writing code and ever want it to be seen or shared, GitHub is the place to do it.
The beautiful thing? You can code a whole data science pipeline, write about your thought process and share a step-by-step tutorial all on GitHub, for free.
Use your readme.md files to yours and others advantage. Don’t just post code, describe what it does and why it’s important. This is something I can still work on.
LinkedIn is the new resume. If you spend 8-hours crafting a beautiful resume, you should at least do the same on your LinkedIn profile.
At a bare minimum, fill out the required fields, write a description about yourself and include a headline.
A good headline should tell someone what you do or at least intrigue them in under 10 words.
What would you like someone to think of when they see your profile for the first time?
Unlike GitHub, you can’t share code directly to LinkedIn but you can share text, photos, videos and even write articles which live on your profile. There’s no secret formula for posting. I got started with a 30-day LinkedIn challenge. Every day for a month, I posted something and it was a great way to start building momentum on the platform.
The beautiful thing about LinkedIn is there are people like you all over it. Everyone is in some way looking to improve themselves and help others. Don’t be afraid to reach out to someone in the industry you’d like to get into and ask for their advice. If you message 47 people and only get 2 replies, don’t take it as a bad thing. Everyone has their own life as complex and interesting as yours.
LinkedIn is one of the first places people go if they’re looking for someone to hire. If you want a role, have yours ready.
Writing not your thing? Better on camera? Make a video sharing your latest project and share it on LinkedIn.
Learned a new thing you think would be helpful for others to know? Create a YouTube tutorial for others.
The internet has given us a plethora of ways to showcase our work. The only way to find the best one for you is to try a few out and see what sticks.
This goes hand in hand with adaptability. You can’t adapt if you’re not eager to learn.
Technology is improving every day. What was state of the art last month won’t be state of the art next month.
“What do you look for in interns?” I asked.
“How quickly they disregard their old methods for learning something new and better.”
Have you ever tried to have a conversation with someone who was right no matter what?
It’s not fun.
Continual learning means eventually you will be wrong. There’s nothing wrong with being wrong. The only time it becomes wrong is when you don’t accept you’re wrong and move forward. Instead, you’re stuck in an old way of doing things.
This was hard for me to accept when I started. I thought I could do everything. I thought it should come easy. Wrong.
Learning something new is hard. Accepting there will always be something new to learn doesn’t make it any easier it means there are no surprises.
You’ve got the skills, you can adapt like a thermophile, you’re showcasing your work better than most art galleries and you’re more eager to learn than 8-year-old with their first iPad.
Sweet, but what do you do now?
Apply. Work. Wait. Repeat.
Showing your work may lead to opportunities but it’s likely you’ll have to do much of the outreach yourself. The traditional application through a careers page is a good place to start. After here, if you really want to stand out, try writing an article about something to do with the role. Send it along with your application. Or make a website.
Be creative. You’re more than a one-page resume.
You’ve made your applications. No one is reading your articles. You haven’t heard back from anyone. It’s been three weeks. All hope is lost.
Not so fast.
It’s time to work. It’s time to figure out where you’ve been going wrong. Take stock of where you’ve come. What’s worked, what hasn’t?
Perhaps the one thing you’re missing is patience. Finding a role is difficult. You should treat it with as much care as starting a new relationship.
Not everyone you ask would be happy to go on a date with you, it’s the same with applying for roles.
Keep applying. Keep learning.
As an up and coming data scientist, you’re looking for ways to use your skills to help make sense of the world. You’re starting a new project but you’ve only got a sample size of one. What’s the first thing you do?
“Look for more samples.”
You’ve just read this article. It’s given you a few things to think about. Are they the best?
I can’t answer that. But I can give you somewhere else to look. I found an article yesterday which contradicts half the things in this one.
The internet has given us access to an abundance of information, all of which you can remix with your own ideologies into your own unique system.
The tips I’ve put together here are from the Chief Data Scientist of realestate.com.au. Their goal is to match people interested in property to the best properties and agents. The company you apply to may have different goals and thus different outlooks on how they hire.
One thing is for sure, finding a role is hard. There’s no set path.
We didn’t mention it during the discussion but I’m going to finish with a point of my own.
Whatever role you’re after, go for it. Apply and apply again. And if there is no role for you, create your own.
You’re more capable than you think.
If you’re more of a visual learner like me, there’s a video version of this article on my YouTube channel. And a bunch more videos documenting my journey into the world of data.
Originally published at mrdbourke.com.