How I Became a Data Scientist Before I Joined LinkedIn
How does someone get a dream job as a data scientist for the first time?
My dream job was to work in data science at LinkedIn.
However, I didn't have any formal data science education nor a previous data scientist job title. Data science was an emerging new field at the time. It wasn't as commonly taught in schools as it is now nor available in most companies.
Before I was hired by LinkedIn, I still needed to prove that I could do the work required.
I'm grateful that I was ultimately hired by LinkedIn. I later became the head of a LinkedIn data science team of 14 data scientists to acquire and grow our customers for our multi-billion dollar online advertising Business.
I'll share with you how I got started with data science in the first place.
Plan For Career Transformations
It's challenging to get your first data scientist job if you're not already working as a data scientist. Perhaps you're graduating from school soon. Or perhaps you've been working for a while as an engineer, a financial analyst, or a scientific researcher, but want a career transformation into data science.
The extra challenge is that career paths are no longer linear. You can expect to refresh your career and skills every few years. In fact, the World Economic Forum cites US Bureau of Labor Statistics that American adults have an average of 12 jobs by the time they're at age 55. The World Economic Forum 2023 Future of Jobs report also predicts that 44% of workers core skills are expected to change in the next 5 years, with the fastest growing jobs to include data skills in AI and analytics.
When I surveyed my LinkedIn network, only 50% of the respondents indicated they're working in jobs which are still directly related to their academic majors. Many data scientists that I know, and even those that I hired, have been like me in starting off with a different academic and professional background. Yes, people can aspire and successfully take steps for career transformations!
Career Transformations Require Real-Life Experience
People that I mentor and coach often ask me "Jimmy, how did you become a data scientist?"
They see that I didn't start off as a data scientist, but eventually became a manager of a data science team at LinkedIn. They want to know what enabled me to transform my career to data science. They themselves are often looking to architect their own career transformations into a data science or AI role too.
To get those career opportunities, I advise people that they first need the following three fundamental things that every professional should be investing in their career anyway, for data science or not:
- Skills: demonstrate the required technical and non-technical skills as a T-shaped or Pi-shaped professional [Learn more]
- Network: grow a strong network of people who can help you with the info, resources, and introductions for new opportunities [Learn more]
- Professional Brand: build your reputation for people to trust you [Learn more]
Growing the above personal assets will help anyone's career be ready to grow and pivot.
For a role in data science or AI or other technical field, people also need to demonstrate successful projects to build trust with employers. It's not enough to have just read about data science, but aspiring data scientists need to show real-life hands-on experience in data science too.
How to Get Hands-On Experience For Data Science
If you're not already a data scientist, then you'll need to complete a meaningful data science project any way you can. Here are three ideas:
- Volunteer to do data science work. For example, check out the Statistics Without Borders group which coordinates volunteers with non-profit NGOs on data projects.
- Complete your internship, or school's business consulting project, or a special data science "bootcamp" program. Get experience working with real-life messy data and interacting with real business stakeholders.
- Propose and complete a data science project at your current employer. For example, I once sponsored an HR analyst to work on a side project on my data science team for her to get some hands-on experience, similar to job shadowing.
For my own initial transition into data science, I was fortunate to use the 3rd option with one of my previous employers.
In the previous company, I had opportunities to explore and propose new projects. I was transitioning from my role as the manager of the data warehouse and business intelligence team, where I had already learned SQL and Python and Linux. Although I was interested in data science, the company wasn't quite ready for data science at that point.
However, I was still able to propose and complete two pilot data science projects there, which provided essential learning opportunities for me.
My First Data Science Project: Sales Data
I remember my first data science project was to classify sales orders in Africa.
My company recorded customer sales orders globally in the online ERP order transaction system. Some of the sales orders were direct sales to customers, which should have credited our internal sales team. However, other African sales orders were supposed to be through an external channel sales partner, which meant the sales should have credited the external partner instead.
These internal and external sales people were dependent on accurate data for accurate sales compensation.
Unfortunately, the company didn't have a way to record the proper sales credit into the ERP database. I don't remember the reason why. The techno-functional business systems analyst couldn't develop a process to record the correct sales credit for the external channel partner. The business couldn't describe consistent rules needed for classifying the sales orders.
The workaround was that the sales team needed to review every sales order coming in, and then manually classify whether the internal sales rep or the external channel partner would get the sales commission.
No one liked doing the manual data classification work.
The sales director approached me for a solution, after hearing about my reputation from his financial controller.
Proposed Data Science Techniques
When I heard about this business problem, right away I felt that this was a problem that might be solvable with a data science classification model.
Data science was new to me and the company at the time. No one at the company was doing any data science yet. In keeping up with industry trends, I was at least aware enough of the techniques used by data scientists in other businesses.
I gave the sales director my plan. As he wasn't technical, I simply told him that I would set up a way to "predict" how each sales order should be classified for sales credit. He waited for what I could deliver.
We didn't have any data mining tools at the time. I ended up researching and settling on using the free Orange Data Mining software from the University of Ljubljana in Slovenia. As a data science newbie, I appreciated that this software had a rich Windows graphical interface where data mining and machine learning data pipelines could be built via the drag-and-drop interface.
The early versions of the software were buggy at times, and the GUI crashed on my Windows computer when the dataset sizes got too big. However, I fondly remember the software was my entry point to learn about different data science and machine learning techniques.
The free software could do all sorts of data science tasks. It could do exploratory data analyses, classifications, and regressions. It could do unsupervised learning including hierarchical clustering and k-Means clustering. It could run different algorithms including Naive-Bayes, SVM, kNN, and neural networks.
Orange was a free no-code data mining system that was ideal for my initial hands-on experience with data science.
The Successful Solution for Sales
With the software, I found that I was able to build a predictive model to classify the sales orders with reasonable accuracy.
For the training set, I had used the historical data from the manual classifications. The data features were all from the sales order records. I ended up selecting the SVM algorithm to build the best model for this business case. I operationalized the model by building a Python script for the data pipeline including incremental retraining of the model with new data.
In addition, I built a Python-based web app for the sales team to confirm the predicted channel partner sales orders, and to review any records with borderline prediction scores. The new confirmations from the sales team were fed back to the SVM model as closed-loop feedback that improved the model accuracy over time.
The sales director was pleased with the results.
This first data science project at the company saved his team time from needing to carefully review every incoming sales order. I was happy to see that the data science techniques actually did work. Data science and AI can provide shortcuts to manual labor in the business.
The sales director subsequently wrote a recommendation for me, which was helpful in gaining trust from future employers.
My Second Data Science Project: Employee Data
After that sales classification project, I did a second data science project at the company to predict employee turnover using simple internally-available employee directory data.
I monitored the results of my model's predictions over a few months, and was encouraged by the model's ability to predict employee turnover.
I shared the results with company leaders, but unfortunately the results were not actionable by the business. We could not intervene with employees whom the model predicted would leave the company.
These are some of the lessons that I later draw upon when I advise data scientists in the people analytics function of HR departments in companies.
Stay Agile For Career Changes
I gained valuable experiences from both of these early data science projects. I'm grateful to have had the opportunity to propose and complete those projects at the previous company when I was not officially a data scientist yet.
Those data science project experiences were crucial for me when I later successfully interviewed for the LinkedIn business analytics (data science) team.
For people who are interested in transitioning to a data science or AI role, find any way to complete a successful data science project. To get accepted into a new field or industry, you'll need to build trust that you already have the needed experience for that dream job.
I've learned from my own career journey as well as from those whom I've coached. In a world where the workplace changes so fast, I do see people take steps to successfully pivot into new Careers like I have, whether for data science or not. In every case, it's vital to keep growing with continuous learning and agility to future-proof your career.
Enjoyed Reading This Story?
Follow me and get notified when I publish a new story: subscribe on Medium for free
