Reading a job description in tech can often leave you feeling more confused about what the company is looking for than enlightened. I've found in Data Science, these job descriptions tend to be vaguer and sometimes don't even align with the company's needs. Unfortunately, in most companies, the Engineering and Data Science departments are far removed from HR and those creating job descriptions. What this means for you is that you can't always trust a job description to do what it's supposed to, describe the job. If you take the responsibilities listed with a grain of salt, it's easier to throw your hat in the ring and get more details about the day-to-day from the team later. No, this really isn't the most efficient way to do things and it certainly makes things harder for you, the candidate, but it takes widespread cultural change and hiring shifts to better align job descriptions with actual job duties.
Think of a job description like a kid's Christmas wish list. The same way a kid writes out everything they could possibly imagine they want to play with, but it'll still be a "successful" holiday if they only get a few things on their list. That's how many companies see a job description. Ideally, they could get someone to do all of the things they listed, but they know they won't get someone who can. I know sometimes we get hung up on the list of tools and how proficient we are in them, but that's rarely the most important thing to a company. Good companies know it's easy to train a motivated employee on a new tool than it is to make someone who's completely uninterested care about solving the business problem.
The first step in understanding these job descriptions better is to know you're not going to check every box and that's okay. Now that we have reset our expectations, let's dive into some job roles. These descriptions are taken directly from job listings as of 04/29/2020.
Most roles of the Analyst variety work with tools like Excel, SQL, and Tableau/Looker. In organizations, these roles are very close to the business and often require reporting to decision-makers. In a lot of organizations, a great way to get started in Big Data is as a Data Analyst. These roles are a good place to start for those who don't have a technical background, but it's important to find roles where analysts use statistically rigorous methods to inform decision-makers.
You should be able to analyze data, create reports on them, provide good visualizations, present these reports to decision-makers. These jobs typically ask for a Bachelor's degree. If you find yourself in one of these roles it's a steep climb to "Data Scientist", but you'll typically need to work on your mastery of SQL, statistics, and coding and statistical analysis in Python or R.
Data Science roles get tricky so we can narrow them down by their strengths. I think the best way to explain this is to use the 9 main strengths outlined in this Data Science Central article.
I would argue that instead of the last category being "those strong in one or more of the above" it should be those strong in a business domain/vertical. That's how I got started in Data Science. I had way more marketing experience than I had coding or math. Once I learned those I was able to branch out and work with other types of data, as well as increase my skills in the other 8 categories.
I like job desroptions like the one below that include what you're going to do and what skills they think you "need to have" to do them. Pay more attention to the top half than the bottom. If you can do those tasks regardless of the tools you use, apply for the job.
You should be able to retrieve and transform your data, perform deep analysis, model and evaluate data models, as well as productionize and present on your findings. Data Scientist jobs typically ask for a Master's or Ph.D. in a quantitative field, but that doesn't mean you have to have one. The Ph.D. is hardly a necessity for companies unless the job title is Research Scientist or you're doing very custom algorithm-based work. If you don't have an MS in a quant field that's okay too! There are plenty of paths into Data Science including starting as an Analyst, taking a boot camp, or self-teaching.
These roles typically require a very strong software engineering background and most interview questions revolve around productionizing models and engineering concepts. These roles are sometimes more similar to software engineering roles and can sit with SWEs rather than Data Scientists and Analysts. People in these roles work closely with Data Scientists and sometimes Data Engineers to develop machine learning pipelines and models.
You can write production-ready code, develop custom models based on business needs, and optimize production models. These jobs are where I've seen folks with a background in software development start. If that's you, you'll typically need to learn machine learning frameworks like Tensorflow, PyTorch, Keras, and Spark. If this is your target position work on your engineering skills and building tools that incorporate machine learning.
Company Size plays a large part in how much work you'll have on your plate. This is less obvious from a job description, but that's why it's extremely important to ask interviewers how large of a Data Science Department the company has. If you're applying for a machine learning engineer role at a startup you should expect to do some of the ETL and exploratory analysis in addition to your productionizing. Here are a few ways the size of your company's data science group can affect your duties.
Xtra Small (1 Data Scientist)
If you’re the only data scientist in an organization you can expect to do most of the heavy lifting when it comes to project design, direction, and steering ideas. You’ll probably have only one data engineer if at all so you need to be comfortable performing your own ETL and data cleaning. Strong project management skills are important as you’ll be in control of the process from beginning to end. These roles are good for strong Data Scientists with a lot of experience because the unknown unknowns are large for those early in their career.
If you’ve got a few folks to help you out, you can expect to work closely with them to talk through modeling, pair program with, and outline priorities. You’ll want to have a good working relationship on your team and understand your relative roles. Some people may be more averse to speaking to execs and stakeholders while some want to step up in that role. You may have goals assigned by leaders in Engineering or Analytics that you work closely with your team to divvy up.
Smedium (5 - 10)
With some close comrades, it’s easier to take a more specialized role at your company. At Smedium DS departments you may see people start to break into specializations like analysts, ML engineers, and NLP or Computer Vision specialists. At this size, it becomes easier to follow some of your interests where it aligns with company goals. If you’re highly interested in ML in embedded systems, you may have more leeway to do more embedded projects while still fulfilling the main responsibilities of your role.
Mid Size (10-20)
At this size, you may see some smaller teams emerge within your team. Some organizations break their DS departments up by both job function and type of data they work on. Others form product teams that can include people with different sorts of data roles. You may also see data engineering or data warehousing split out as their department grows to sustain analytical growth. One of the biggest challenges here is finding your role in the department. New hires should to find gaps in the groups’ knowledge and fill them.
Departments of this size tend to leverage project managers and full data engineering teams to support analytics. What’s nice is you may have more direction and resources at the start of new projects, but sometimes work can just feel like you’re doing only assigned tasks. Here you want to separate yourself as a specialist or a high-performer on your team. Leverage that each team member is probably more versed in one topic and set up time to learn from them one on one. You have a huge opportunity to help other folks on your team with data cleaning or take charge of new project ideas without worrying about abandoning priorities.
At a large Data Science department, you can expect data or models to be part of the product or the company as a whole is large, a la Google. Here you can expect your work to be more siloed to the product team you’re on or you may have few projects outside your domain (ml modeling, data engineering, etc). These roles are good because you have a lot more resources available and you may have more mentors or mentors with more time to devote to helping you grow. You may only work with one subtype of data, or you may be customer-facing and sell the product with your knowledge of what it’s capabilities are.
❓How to know if you should apply for a job?
If you meet 25% of the role's responsibilities, go ahead and apply!
Finding a job in Data Science isn't just about checking off boxes on a list of requirements, you need to learn how to spot your transferrable skills and communicate that to hiring managers. There are so many valuable skills for Data Science I got in my Social Media jobs! Once I got good at articulating how even my non-technical skills translated to me being BETTER suited for the job than others, landing roles was a breeze.
I'll be talking more in-depth on how to do this on my Data Science Job Description webinar. Right now it's in the works so sign up for my newsletter to hear first when it drops. I'll be going over my own job description and tips for deciding which jobs to apply to.
Interested in Data Science, but you aren't sure where to start? I can help you build a foundational knowledge about changing careers in my free Getting Started in Data Science webinar.