From Build a Career in Data Science by Emily Robinson and Jacqueline Nolis
You’ve got the skills, and you’ve got the portfolio – all you’re missing is the data science job!
In this article, we’ll focus on how to look for data science jobs. You’ll first learn all the places where you can find jobs, making sure you won’t unknowingly narrow your options. We’ll cover how to decode these descriptions to find out what skills you need (spoiler: it’s not all of them) and what the job might be like. Finally, you’ll learn how to choose which ones are best suited for you, using your knowledge about data science skills and company archetypes.
Before worrying about crafting the “perfect” resume and cover letter, you need to know where to send them. Job boards like LinkedIn, Indeed, and Glassdoor are a good place to start your search. It’s worth looking at more than one website, because not all companies post on each one. If you’re part of an underrepresented group in tech, you should also look for job sites targeted specifically at you, such as POCIT and Tech Ladies, for people of color and women in technology respectively. The type of job you’re applying to might also influence where you look. There are job boards for specific types of companies like start-ups (AngelList) and technology (Dice).
Make sure to browse widely. Data science jobs go by many names besides data scientist. Different companies use different names for similar roles, and some are even are changing what their titles mean. All the people who were data analysts one year might be data scientists the next, with no change in responsibility!
Some examples of titles you might encounter include:
- Data analyst: This is more often a junior position and can be a great way to start in the field if you don’t have a STEM degree and haven’t done any data analysis for a company before. As we’ll discuss later in this article, you want to be extra careful with data analyst positions to make sure that the role involves programming and statistics or machine learning.
- Quantitative, product, research or other non-data analyst: These roles have even more diversity than data analyst in terms of your responsibilities. You may be doing exactly the same type of work as “data scientists” at other companies, or you may be spending your days with legacy Excel spreadsheets.
- Machine learning engineer: As implied by the title, these focus on the machine learning part of data science. They usually ask for a strong engineering background—if you have a degree in computer science or have been working as a software engineer, this could be a great role.
- Research scientist: These positions often require a PhD, though there might be some negotiation room if you have a master’s in computer science, statistics, or a closely related field.
When you’re first starting your search, try searching for “data” on one of these job boards and spending an hour reading through job posts. This gives you a better idea of what industries are represented in your area and what types of positions are open. You’ll pick up on patterns that let you skim through new listings more quickly. Finding jobs which are a good match for you, rather than all the available jobs, narrows the field down to a manageable number. Don’t worry too much about the title—use the description to evaluate fit.
Be extremely cautious about thinking of job-hunting as a numbers game. If you’re looking in a big tech city like New York or San Francisco or in multiple cities, you’ll find hundreds of jobs listed. Checking job boards can quickly become an obsession. It’s an easy way to feel productive: “I read through 70 job descriptions today!” And like Twitter and Facebook, checking constantly for updates can be addictive. But checking more than every three to five days doesn’t generally add value. Although checking only once a month could mean you miss out on a good opportunity; we know of no company which has filled a position within two days of posting on a job board.
If you have specific companies you’re interested in, check out their careers page. As you should search for multiple job titles, check different departments—some companies may put data science under finance, engineering, or other departments, and if you hadn’t checked there, you wouldn’t have found them.
Once you start reading job descriptions, it can seem like data science job postings fall into one of two categories. One is a business analyst position where you’ll use business intelligence tools like Excel and Tableau, with maybe a little SQL, but generally you won’t code. If you want to grow your coding skills, machine learning toolbox, or statistics and data engineering knowledge, these aren’t a good fit.
A common listing at the other extreme describes a unicorn—a PhD in Computer Science who has also worked for 5+ years as a data scientist and who is an expert in cutting-edge statistics, deep learning, and communicating with business partners—and lists a huge range of responsibilities, running from doing production-level machine learning to creating dashboards to running A/B tests. These types of job descriptions usually mean the company doesn’t know what they’re looking for, and they expect a data scientist to come and solve all their problems without any support.
Don’t worry though—we promise there are more than these two types. The better way of thinking about these jobs is in terms of experience. Are they looking for someone to build a department of their own with no data pipeline infrastructure in place, or are they looking for a fifth member of a currently productive data science team where they hope that person contributes immediately, but isn’t expected to be an expert in data manipulation, business communication, and software development all at once? To do this, you need to take a job description and figure out what they’re looking for. Imagine looking at cat adoption listing and the cat Honeydew Melon is stated to “like asking about your day.” You’d want to realize that means she will constantly meow for attention, which could be bad for your home. By knowing how to read between the lines, you can make sure you’ll be applying for the right jobs.
The first thing to keep in mind is that job descriptions are generally wish lists with some flexibility. If you meet 60% of the requirements (e.g. you’re a year short of their required work experience or haven’t worked with one component of their tech stack), but are otherwise a good fit, you should still apply. Definitely don’t worry too much about the “plusses” or “nice to haves.” Additionally, years of work experience requirements are a proxy for having the necessary skills—if you coded in grad school, that could count. That being said, applying to a post for a senior data scientist that requires five years of work experience as a Data Scientist, proficiency in Spark and Hadoop, and experience deploying machine learning models into production probably isn’t the best use of your time if you’re an aspiring data scientist; they’re looking for a different level of experience and qualifications.
Many data scientist jobs list a degree in a “quantitative discipline,” meaning fields like Statistics, Engineering, Computer Science, or Economics, as a requirement. If you don’t have one, can you still apply for them? Generally, yes; if you took classes in those areas (including in a bootcamp or online), you can emphasize that. If you built a portfolio and wrote blog posts, you can show those to employers as evidence you can do the work.
One complication in data science postings is that there are different words for the same things. Machine learning and statistics are infamous for this. One company may ask for experience in regression or classification, another in supervised learning, but these are overall the same thing. The same goes for A/B testing, online experimentation, and randomized control trials. If there is a term you’re not familiar with, google it – you may find you’ve done it under a different name! If you haven’t worked with a specific technology they’re referencing, see if you’ve done something similar. For example, if they cite AWS (Amazon Web Services), and you’ve worked with Azure or Google Cloud, you have the skill of working with cloud-based technology.
The other benefit of knowing how to decode a job description is the ability to detect red flags. No company is going to straight up say that they are bad to work for. The earlier you recognize a likely bad work situation, the better, and you’ll want to start looking for any warning signs in the job description.
Watching for Red Flags
Finding a job is a two-way street. It can feel during this process that companies have all the power and you need to prove you’re deserving. But you, yes you, can also be selective about where you work. Ending up in a toxic workplace or a mind-numbingly boring job is a hard situation. Although you won’t always be able to tell if this is the case from a job description, there are a few warning signs to watch out for.
The first is if there’s no description of the company or job itself, only a list of requirements. Those organizations have forgotten it’s a two-sided process and aren’t thinking about you. It can also mean they’re buying into the data science hype and want to have a team of data scientists, without setting anything up to work productively.
A second warning sign is the aforementioned “unicorn” description. Although that was an extreme example, you should be careful of any job description that describes two or three of the job types (decision scientist, business intelligence, and machine learning) as primary responsibilities. Although it’s normal to be expected to have base competency in each, no person is going to be able to fill all those roles at an expert level. Even if someone could, they wouldn’t have time to do it all.
Finally, look for mismatches between the requirements and the description of the position. Are they asking for experience with deep learning but the job functions are making dashboards, communicating with stakeholders, and running experiments? If it is, they may want someone who can use the hottest tool or who is a “prestigious” data scientist, maybe someone with a Stanford PhD in AI, when they can’t use that specialized knowledge.
Setting your expectations
Although you should have standards for a potential job, you don’t want to demand perfection. Aspiring data scientists sometimes see their path broken down this way: “Step 1-98: Learn Python, R, deep learning, Bayesian statistics, cloud computing, A/B testing, D3. Step 99: Get a data science job. Step 100: Profit.” Although this is an exaggeration, part of the data science hype is the idealization of what it’s like to work in the field. After all, data scientist is “the best job in America” (according to Glassdoor), with a six-figure paycheck and high job satisfaction. You might imagine getting to spend every day on the most interesting problems in the field with the smartest colleagues, and the data you need is always accessible, cleaned, and any issues you face is solved immediately by a team of engineers. Your job is exactly as it was described, and you’ll never have to do the parts of data science that interest you less.
Unfortunately, this is a fantasy. You don’t need to know everything before getting into the field because companies aren’t going to be perfect unicorns either. This article doesn’t end with you getting a data science job. Although it’s a great accomplishment and you should be proud, data science is a field where you’ll always be learning. Models fail, workplace politics scrap the work you’ve been doing for the past month, or you’ll spend weeks working with engineers and product managers to collect the data you need.
It’s easy to idealize companies that are well-known, either generally or for data science. Maybe you went to a talk and one of their employees blew you away. Maybe you’ve been following their blog for months and know they’re on the cutting-edge of the field. Maybe you read an article about how they have nap pods, gourmet meals, and lots of friendly office dogs. But whatever has attracted you has likely interested other aspiring data scientists as well; most of these companies get hundreds of applications for an open position and can set the bar higher than even what is needed to do the job. In any case, the work you read about might be in a totally different division and this position may be uninteresting.
Even with realistic expectations, you’ll likely not end up in your dream job for your first data science role. It’s easier to transition within your field or bring data science into your current role; even if you’re looking to eventually leave your domain, you may need to start out by moving to a position where you can use your other skills. That doesn’t mean you shouldn’t have certain requirements and preferences, but it does mean you’ll want to have some flexibility. It’s normal to switch jobs in tech even after a year or two; you’re not signing yourself up for the next fifteen years. You can’t know exactly what you want before you’re even in the field, and you’ll learn even from bad jobs; don’t stress too much.
Leveraging your network
You’re not limited to looking at job boards. Many data science meetups have time at the beginning where people can announce if they’re hiring. Conferences also might have job fairs or at least booths where sponsoring companies may be looking for new hires. Go up and talk to these people; it’s part of their job, and even if their current openings aren’t a good fit they maybe be able to give you some good advice or suggest other places to look.
You may also meet another attendee who works in the company or sub-industry you’re interested in. You can ask if they have time for an informational interview to learn more about their field. An informational interview isn’t (or rather shouldn’t be) a passive-aggressive way of looking for a referral—instead, it’ a great way to get a look inside a company and get advice from someone who is in the field. We don’t recommend asking people you met that day to refer you. This is a strong ask for someone they don’t know, and no one likes being used. If they tell you about an opening at their company and say they can refer you, this is a great bonus, but you’ll gain a lot even if they don’t.
If you’re able to be public about your job search, post on social media that you’re looking and ask if anyone has any leads. Even though you might not have a strong data science network yet, hopefully you have friends and former classmates and colleagues who might know of positions within their companies. This generally works better on social media where people connected to data science tend to congregate, such as LinkedIn or Twitter, but even social media networks like Facebook may have connections with opportunities.
Although social media posts can be great for getting a broad range of people to see that you’re looking for opportunities, targeted approaches can be helpful too. If you happen to know people who might have leads for you, sending direct messages isn’t uncommon and won’t be frowned upon. If you haven’t met the person before but they share a relation to you, like you studied the same major in the same university, it can be worth it to try.
Lastly, it’s common early in your career to feel like you have no network yet—networks seem to be held by people who already have data science jobs! The solution isn’t only to network when you’re looking for a job, but long before it. The more you can get out of your comfort zone and talk to the people around you at places like conferences, meetups, academic institutions, BBQs, etc. the more prepared you’ll be when it’s time to look for a job.
You may find yourself in a data science adjacent job. An unusual, but often effective, method of learning data science is to start doing more and more data science work as part of your current job. Maybe you’re in business intelligence, set up databases that data scientists use, and could start doing queries. Maybe you’re a business person who takes the data science reports and already adds a business spin, and could start also adding your own graphs. Maybe you work in finance making spreadsheets that you could move into R or Python.
For example, consider “Alex,” a person who has been working for several years in the market research department running surveys on customers and using a market research GUI (Graphical User Interface) to aggregate the survey results. Alex has a background in sociology and did a small amount of programming during their undergrad. Alex frequently works with the data science department—Alex passes them survey data and helps the data scientists understand it to use it in models. Over time, Alex starts to do a bit of work for the data science team. A bit of feature extraction in R here, a bit of creating visualizations there, and soon the data science team is relying more and more on Alex. During this time Alex improves his programming and data science skills, and after a year Alex fully joins the data science team, leaving market research behind.
Trying to do some data science in your current job is a great method because it’s low risk and has built in motivation. You aren’t trying to do an expensive bootcamp or degree that you have to quit your job for, you’re trying to add a little data science work where you can. And the fact that you are doing data science in your own job is motivating because the work you’ll do is valuable to others. Over time you can do more data science work until eventually it’s all you do, rather than trying to do an educational program and suddenly switch all at once.
Our former market researcher and now data scientist Alex had a number of things going for them in our example. Alex already had existing relationships with the data science department to provide mentorship. Alex also understood the basics of programming and data visualization. Alex was self-motivated enough to learn data science techniques within their job. The data science department was able to provide Alex with small projects that Alex could tackle, and over time these grew to enable Alex becoming a data scientist. When trying to do more data science work in your company, look for places where you can find small data science projects and people to help you with them. Something as simple as creating a report, or automating an existing one, can teach you a lot of data science.
One important note when taking this path: never become a burden for someone else. That could be obvious, e.g. direct burdens might be repeatedly asking people to send you cleaned datasets. Or it could be less obvious things like constantly asking someone to review work you’ve done. You can also be an accidental burden by adding new tools to your team, i.e. if you’re in finance and everyone uses Excel except for you who now uses R, you’ve made managing your team more complicated. Even the task of asking a data scientist for work you can help with can be a burden. Be thoughtful as you learn these skills that you aren’t creating issues for other people.
What you say: “I am happy to help in any way I can, let me know how! Thanks!”
What you think they hear: “I’m a person who is eager to work for you, and you can hand me that exciting but simple project you’ve held onto and I’ll do it for you!”
What they hear: “Hi! I want to be helpful but I’ve no idea what your needs are. I also don’t know what my skills are relative to your workload; good luck finding a task for me to do. Also, if you do somehow find a task this is perfect for me, you’ll probably have to review it a bunch of times before it’s good, all of this taking away from your already minimal free hours. Thanks!”
To make this path work there are a few key strategies to employ:
- Be proactive – the more you can do work before people ask for it, the more you’ll become independent and less of a burden. For instance, the data science team may have a task, like labeling data or making a simple report, which is time-intensive and uninteresting. You can offer to help with that work. Be careful diving in entirely and doing it yourself—you may end up doing it in a way that instead of providing value provides the team the opportunity to redo your work. But if you can get it started and then get their input, it’s possible you can save them a lot of time.
- Pick off new skills one at a time – don’t try to become a data scientist all at once. Find a single skill you want to learn through work and then learn it. For instance, you may want to learn how to make reports with R because the data science team does that all the time. By finding a small project to help the team with, you can pick up the skill and add it to your tool box. From there you can learn a different data science skill and gradually learn enough that way.
- Be clear with your intentions – it’ll be pretty obvious fairly quickly that you’re trying to pick off extra work to learn to be a data scientist. By being proactive and letting the data science team know that you’re interested in learning more, the team will be able to plan around having you help. It’ll also make the team more understanding of your inexperience, because they were learning and new once too.
- Avoid being pushy – helping a person become a data scientist is an immense amount of work, and data science teams are often already overworked. If you find that the team doesn’t have the time or bandwidth to help you, don’t take it personally. Although it’s okay to occasionally check in if you think they’ve been out of touch, if you are too persistent with your requests the team will quickly become uncomfortable. They’ll view you less as a potential resource and more as a nuisance.
Learning on the job summary
Learning on the job can be an effective way to become a data scientist, provided you have a job where you can apply data science skills and people around who can mentor you. If those things align then this is a great route, but for many these things are not in place. If you do think this is a viable route for you, we highly recommend taking it—it isn’t often that jobs allow for this. Take the opportunity if you have it.
Stay tuned for part 4, in which we will discuss what to expect once you land a job.
That’s all for now. If you want to learn more about the book, check it out on our browser-based liveBook reader here.