I’m a Data Scientist, Not Just The Tiny Hands that Crunch your Data
Not everyone "gets" the role of the Data Scientist -- including management. While there can be frustrating aspects of being a data scientist, there are effective ways to go about fixing them.
By Ahmed Besbes, AI Engineer // Blogger // Runner.
This comes out as a personal observation, but I’m sure that many of you will share the same feeling upon reading this post.
I’m a data scientist, and I like my job because I think it covers various interdependent domains that make it rich and stimulating. However, I sometimes have to deal with people who don’t exactly understand this role in the organization nor the field in general. This, quite frankly, is what makes things a little bit frustrating for me and also for a lot of people I know.
Before you keep reading, I should mention that I don’t aim to discourage anyone from aspiring to this role. I’m only stating some negative aspects that occur in the industry in general and the possible solutions for avoiding them.
Some people don’t exactly understand what you do … and don’t even bother to explain!
In principle, this is fine. I don’t understand what most other people do either. What I do not get, however, is the total lack of interest and curiosity of some parties in learning about what you do while helping them. I don’t mean they should get every small algorithmic detail of your neural network, for example, but at least, they should get to know your approach, your way of solving the problem. Sometimes, it’s as if you were commissioned with the painful, dirty task that no one cares about.
Some project managers take zero interest in what you’re doing unless you’re done doing it. I think these fellas bring management to a whole new level.
Oh! You’re a data scientist? You must be really good with the numbers. Why don’t you have a look at my files and crunch the data? I hear your “python” thing can pop out the magic real quick. Here, go play with my files and come see me when you’re done.
— What to do?
To make everybody on the same page, one solution is to provide training and awareness to the teams who have no technical background. This goes through internal workshops, certifications, or MOOC subscriptions in broad technical topics such as introductory lectures to machine learning, deep learning, or NLP. When building knowledge in these areas, teammates become proactive and more engaged in the building process. Project managers also become aware of the challenges.
Data scientists are still considered marketing tools to pimp proposals
Well, this worked quite well ten years ago when the field started to emerge, and the words Hadoop and Spark were all over the place. You could stack all the buzzwords you know and hope for a big check (and it worked!).
This isn’t 2010 anymore. Companies now pay close attention to what you’re willing to sell. They know the market, the competitors, and the challenges. They’ve scanned nearly everything thoroughly. They also know what’s feasible and what’s not. If you don’t stand out of the crowd and are not clear enough about your value proposition and the technical expertise that your data science team can bring, you’re most likely to lose the deal.
Of course, despite all of this, there is always some ballsy guy in a suit to make this kind of inspirational statement:
Let’s throw a little bit of data science here and there to beef up our pitch and make the client pay a buttload of money!
Isn’t that beautiful?
— What to do?
Don’t act as if a data scientist would completely change and disrupt your organization. The market starts to know what the limitations are. Be in line with the market.
You shouldn’t be the tiny hand that doesn’t take enough credit for its work.
We all know this feeling, and it sucks. You bust your a** working hard, and some other guy presents your results and takes all the credit. This is common everywhere and happens even more when you work in a data science team in collaboration with business partners.
If you’re valuable to the team, your colleagues should naturally let you shine in front of the stakeholders. Your voice is then heard and engaged in the decision process.
If you’re feeling, however, that you’re treated like an interchangeable resource or put aside working in the shadow and producing numbers for those who speak, maybe it’s time to rethink your position.
— What to do?
Everyone is important when building a data product. This should not only be a statement that we tell ourselves. It must materialize in our meetings, presentations, and daily relationships.
Data scientists cannot produce insights upon request
Well, as tempting as it sounds, this is not as easy as we think. Just because we’re equipped with the tools doesn’t necessarily mean that you can expect immediate actionable results. This requires building knowledge about the business, forging the right intuitions, and the assumptions. This takes time, and it’s a learning process.
Let’s crunch the data and make it speak.
— What to do?
Accept the fact that a data scientist has to spend a substantial amount of time learning about the business and building his own intuitions about it. This goes through interviewing different actors in the organization, running all sorts of analyses on the data, experimenting, failing, and getting continuous constructive feedback.
If you also want to provide the best conditions for your data science teams, make sure you have, at least, clean data pipelines with clear descriptions.
A data scientist cannot be the go-to person for every data related issue
There’s still a strong misconception about the role of the data scientist. Not only non-technical executives but other colleagues in tech believe that data scientists know their way around Spark, Hadoop, SQL, TensorFlow, NLP, AWS, production-level applications, docker, and more. It’s great to master these tools, but this process takes several years and a lot of experience.
If you’re a data scientist and you’re applying for a company that mentions all of these techy words in one application, double-check the company. It’s possible that it hasn’t a clear vision of its data strategy nor a clear definition of the role it’s hiring for.
We need to fix our data problems. Let’s hire a data scientist.
— What to do?
A data scientist is not always the ultimate solution to your data problems—double-check before hiring. Maybe what you need is a data analyst or a back-end developer. A data scientist is not a ninja that masters everything.
Pro-tip for those who want to build a strong data team
If you want your team to succeed in building whatever you intend to build, make sure you surround yourself with complementary skills.
At the delivery level:
- Data scientiststo build sophisticated machine learning models, draw complex analyses, and formulate business needs in terms of metrics.
- Data Engineers to build, among other things, robust data pipelines so that data is clean and accessible for the data science team at any time
- ML / AI Engineers: this is a new role emerging in the field. I see it as a hybrid profile between a data scientist and a data engineer. In practice, it’s a data scientist who goes beyond modeling and thinks about deployment aspects. Questions he solves, for example, are: how do I make the model scalable? How do I dockerize my application properly? How do I ensure low latency at inference time? etc.
- Front and back-end devs to build web applications that integrate and package the machine learning logic. They deal with code quality, robustness, security, design, stability, building APIs, etc.
A data scientist can find his way around building small web applications but remember that this not his expertise. If you want a professional mobile or web application, hire a team of developers.
At the management level:
- Data science managers: these are the most technical profiles within the management team. They supervise the data science teams and make sure they take the right (modeling) decisions.
- Project managers: They make sure things are on track in terms of deadlines. They spot the blocking issues and interact directly with the business or the client.
- Chier Data Officer (CDO): This is the top management role. His goal is to infuse the culture within the organization, to look for the projects, and build the business.
This is based on a compilation of discussions and several feedbacks coming from friends and colleagues.
Original. Reposted with permission.
Bio: Ahmed Besbes is a data scientist working across many industries, such as financial services, media, and the public sector. Part of Ahmed's work includes crafting, building, and deploying AI applications to answer business issues.
- If I had to start learning Data Science again, how would I do it?
- What every Data Scientist needs to learn from Business Leaders
- Software engineering fundamentals for Data Scientists