20 Feb Hiring data scientists? You’re doing it wrong
We are looking for someone to fill an upcoming gap in our business model. We are not exactly sure what you will be doing, but we are sure our shareholders will love the idea that we have data scientists.
You will report to someone who does not understand what you do, and you will often be met with skepticism when you present your solutions to the management.
Candidates must possess the following:
- B.Sc in Computer Science
- M.Sc/PhD in a quantitative field
- Three to five years of experience in a research-focused position handling back-end frameworks, application development, and web hosting
- A deep knowledge of both neural networks and statistical methods
- Five years of programming experience
- Two to three years of experience in database management
- Five years of experience in our domain (e.g. healthcare, marketing, etc.)
- Advanced knowledge of SQL, Python, R, Matlab, Java, C, C++
- Two to three years of experience publishing research papers in AI, NLP, or computer vision (all three preferred)
- Experience with Spark, Hadoop, and Pig…and Horse
A person that hardly exists
How many people in the world do you think fit this bill? And what number of those people have the soft skills to be customer facing, client facing, management facing, yet analytical, creative, and intelligent? We are asking the wrong things from data scientists and we are looking in the wrong places.
There is no possible way that a data scientist will use all these tools in one company. It’s even less likely that someone knows all these languages. Data science is more about the intelligent use of programming rather than programming itself.
You are not hiring programmers
Stop focusing on degrees and credentials, but on proven real-world experience. The person does not even have to be a formal data scientist. Here is my main goal when I look to bring on a new member to our ML team:
What we want:
- People who are intelligent, explorative, and passionate—instead of experienced
- People with an ability to take cloudy, uncertain ideas and build them into finished products
What we offer:
- An interesting team and strategy and the opportunity to work on cutting-edge technology
- Room to grow and an understanding of what they want to learn and help them get there
- An understanding that everyone in the team is equal
Start with what matters
The most difficult skills to find when it comes to data science are not tied to a degree or specific work experience. It is the explorative nature, the push for optimization, the results mindset, the never-give-up attitude, and the willingness to learn that makes a good data scientist. When I look to hire people in my team, these are the most important skills. Everything else is gravy.
To be sure, degrees and experience can be a good judge of ability, but those come at a high cost. It dramatically shrinks the size of the pool you look to draw from, and it creates a confirmation bias whereby we continue to push for more and more credentials.
Create a job that you would want to take
When I am hiring, I spend hours working on the job description over several days. I build a role that fits the needs of our team, provides great learning opportunities, and offers something exciting. Without that, people will become bored and bitter over time.
It is easy to identify a business need for a specific role, but it is often difficult to define exactly why you need a data scientist and what they will be doing. It is your job to think this through before starting your search.
Managers should write their own job descriptions
If you cannot spend three to five hours to write a proper job description, then you do not need the position enough.
The number of times I have read job descriptions that sounds generically written by HR is shocking. You are often left with no idea what you are doing, what the team is like, or whether you have the 35 skills required to fill the job.
Be more interested in how people think
Since I am not a computer scientist or have a PhD (I am a mechanical engineer), I may be a bit biased. I was taught how to take complex problems and break them down into smaller tasks, just as programmers are. However, I truly believe that it was not until my undergraduate engineering degree that anyone expected me to think. This skill is so underutilized and undervalued it is shocking to me.
Finding people who can think critically, creatively, and intelligently is hard. Understanding how to take data and explore, analyze, and present it in a valuable way is hard. Taking billions of unstructured data points, extracting information, transforming it, augmenting it, and generating actionable data/predictions is hard.
Perhaps, the most valuable skill a data scientist needs is the ability to handle open problems without a defined solution. He/she needs to be able to provide value when the value is not defined—they often need to understand this and get there on their own. He/she needs to be able to choose from the hundreds of potential outcomes to find the one that best fits the problem and business at hand.
A true data scientist:
- Understands the concept of value creation
- Understands business
- Understands how different techniques and models can be used to create value
- Handles complex, open problems methodically and with interest
- Doesn’t need to be a good programmer
Will this make your job hunt easier? Probably not. But at least you will look for real people and the right set of skills.
When you force yourself to look for this type of person, you also force yourself to think about why you need a data scientist in the first place.
This article was first published on Medium.