It all started in February, 2013 on a sunny Sunday morning in Foster City.
As I was sitting at my kitchen table – I started to contemplate about my future career path. You see, I’d been working in predictive analytics and statistical modeling for about 6 years at that point for a SaaS company, but I felt like my growth has slowed and I wasn’t learning new things anymore. I know I needed to make a change. At the time, all I hear and read in the news was this foreign term called “Data Scientist”. It sounded like it was the coolest thing to be a Data Scientist. However, I wasn’t really sure what it meant to be a data scientist; but I felt like I was close to being one or I could get there with some help. The competitive fire in me started to burn and I was going to do whatever it takes to land the job as a data scientist. With that in mind, I fired up my LinkedIn, updated my profile, and paid for a LinkedIn Job seeker subscription (free trial first, of course!).
Fast forward to today, I am happy to say that I’ve been fulfilling my dream, working at LinkedIn as a Staff Data Scientist/Sr. Manager for 3.5 years. How did I get here?
Well, my path to LinkedIn as described by discrete list of events @LinkedIn over a course of 1.5 years are (not necessarily in order):
- worked with 12 different LinkedIn recruiters across 3 different teams
- 1 internal referral
- 10 phone screens
- 4 separate on-site interviews
- meeting with 24 interviewers
- suffering 3 rejections
- 1 job change in between
- 1 Offer letter
- It was quite a journey to say the least, and multiple times I’ve thought to myself that it just wasn’t going to happen for me. When I finally got the offer, I was overjoyed and felt that I’ve proven myself as a data scientist!
Since this was a very unique and transformative job search experience, I’ve summarized my learnings into the 5 step process below for anyone who’s looking to pursue a similar career transformation as a data scientist.
Step 1. Identify the Gaps
For me, I knew I wanted to become a data scientist, but I didn’t really know what that meant or how to become one. What were the requirements? Where to start? I had to find out for myself, and LinkedIn was a great place to start. I searched for data science related job postings at top tech companies and studied the requirements closely.
First part of the job posting I focused on is the basic qualifications. This is pretty cut and dry for the most part: most required quantitative degrees such as Mathematics, Statistics, Computer Science… etc. That I have, check. Then there would be programming requirements such as a statistical programming package like R or Python. Ok, I am quite familiar with R, but haven’t really used Python much at the time. Thirdly, a requirement for database ETL tool such as SQL. Then comes the kicker – “familiarity with HDFS, MapReduce and Pig/Hive”. Ouch. I worked for a tech company but we didn’t use big data technologies at the time. I haven’t had the chance to use any big data tools at work previously. “What the heck is MapReduce?” so i thought at the time.
Lastly, depending on the job, there would be contextual requirements such as familiarity deploying Machine Learning models for ML Engineer / Data Mining, familiarity with A/B Testing and experimentation for Product Data Scientist, or familiarity with SEO/SEM for Ad/ Marketing Data Scientist, etc. etc.
At the end of my research, I now have a pretty good idea of the things that I have that companies are looking for in a data scientist, and some gap areas on my resume that I needed to fill, somehow. I estimated that I’ve got about 70~80% of the mathematical, statistical, and programming skills covered, while significantly lacking in the big data / web analytics 2.0 / internet company skills.
Step 2. Make a Plan
As with most things in life, you should make a plan. Career Transformation is no different. After having mapped out my gaps, the first thing I did was set a timeline. Why have a timeline you ask? Well, I find that if you don’t force yourself to commit to a set date to get things done and hold yourself accountable, the chances are time will slip by and nothing will get done.
How much time should I give myself to achieve the goal of becoming a data scientist? 1 year seems too long and lacked urgency, and 3 months seemed too short to ramp up on a lot of new knowledge. I ended up going with 6 months: by August of 2013, I was going to become a data scientist! (To be honest, there is no right answer to this question; it’s highly context dependent. You need to estimate how much time you can invest per week, and how narrow/wide the gaps are in your resume to arrive at a reasonable timeline.)
And here’s my plan:
- Update my resume
- Study data science on nights and weekends
- Apply to some data science jobs
- Study data science on nights and weekends some more
- Apply to some data science jobs some more
- Study data science on nights and weekends some more …
- Apply to some data science jobs some more …
You get the idea – It was a very simple plan. The reason I decided to apply as soon as I started, knowing that I have gaps and the chance of success was low was because:
- I don’t really know how much knowledge I actually needed to get hired. Also, I can’t possibly learn everything – I just needed to know enough to get hired and I believe I could learn everything else on the job
- I haven’t interviewed for data science positions before and I don’t know what to expect. The best way to learn how to interview is by actually going on interviews
- Hey, you never know, I might get lucky!
The trick here is to fail early and often, with companies that you could join, meaning they are somewhat interesting to you and is a step-up from your current job, but not your top 5 companies you had hoped to join. Save those for when you feel you’re ready. So the day goes,
Interview, Rinse, Repeat. Interview, Rinse, Repeat…
Step 3. Iterative interviewing (and failing gracefully)
A wise friend once told me that he regularly goes on interviews every year even though he wasn’t looking to make a change. I was like why would you waste the time interviewing if you have no plans to make a change? His answer opened my eyes. He told me that “First of all, you should know your market value. If you don’t know what you’re worth, how can you make the right decisions about your career?” Then he said, “Second of all, interviewing is a skill. If you don’t constantly practice and hone this skill, it dulls”. Wow, point well said. In fact, it made so much sense that I felt I had been living under a rock up till this point.
So, there you go, interviewing is a skill that must be practiced and sharpened.
Sticking with the plan, I started applying for jobs as soon as I can. My goal was to learn how to interview for data science positions, starting with almost no knowledge what it entails. Since I’ve decided to take an iterative approach, I knew I would be terrible in the beginning, but as long as I was learning something each time, I would get further along on the journey. With my LinkedIn Premium subscription activated, I know my resume would be promoted when I applied to jobs on LinkedIn, and it worked like a charm. I’d get phone interview requests every 2 out of 3 jobs I applied to within days. Sure enough, I was getting rejected on technical phone screens and onsite interviews for almost all the data science jobs I applied to in the beginning. “We’re going with another direction, but we’d love to keep in touch and keep you on file” They’d say.
However, I knew the process was working and I was getting better at interviewing. As the interview counts racked up, the nerves went away, and I was getting quite familiar with the interview process. I got comfortable being in a stressful environment, and I could also start to notice a pattern in the line of questioning. My pitch improved – I was able to iterate and tested out different ways to delivery my professional summary. Over time, I prepared 2 minute, 5 minute, and 7 minute summaries depending on how much time I had. My preparation was better – having been exposed and quizzed on various types of statistical concepts, probabilities, product sense, data manipulation exercises etc., I now have a sense of what types of questions might come up and started getting better at not only solving these questions, but also explaining how I would solve these questions. You see, a lot of times the interviewer is not only looking for your reasoning and the correct solution, they are also interested in how you interact with a problem. Invariably, innovation leads you down a path where you’d be presented with problems you’ve not seen before. Therefore, they are more interested to understand your thought process and how’d you interact to challenging problems, in order to assess whether you’d be up to the task.
Lastly, a quick word about failing gracefully. The moment when the call comes – your heart immediate starts pounding; you’re perspiring in anticipation; your mind is off wondering about the outcome while you exchange small talk with the recruiter. All of a sudden, you hear “we regret to inform you…” and boom, just like that, if feels like a sledgehammer hits your chest. Suddenly, your mouth is dry and your stomach turns into a knot. You’ve just been turned down, and you feel you are Just. Not. Good. Enough.
Hey, we’ve all been rejected before. It sucks. It’s important to know that it isn’t just about you. The company is looking to fill a very specific role and a person with a set of skills uniquely suited to fill that need and help the company grow. For companies, it’s a match-making problem of talent, skills, and maximizing Return On Investment (ROI). I used to feel compelled to find out exactly why I was turned down. My advice? Politely thank the recruiter and interviewers for the opportunity and move on. Failing gracefully is to learn from the mistakes, don’t dwell on the negatives, and focus your energy on the next opportunity.
Step 4. Skill Acquisition: Information >> Knowledge >> Skill >> Mastery
In the internet age, vast amounts of information bombards our every sense at every waking moment. The way information registers and becomes knowledge has drastically changed for many of us. I believe Skill Acquisition is a fundamental survival instinct for the modern age professional, and that there is a 4 step process from information to knowledge to skill and finally mastery. Since this process warrants a blogpost of its own, let me simply focus on how it relates to filling the gaps on my resume.
So where could I acquire knowledge that I don’t have from classical training?
One word – MOOC.
You see, right around the same time in 2012-2013, online learning started to take off and getting mainstream attention. To my delight, when I started googling for terms like “learn data science”, “top machine learning algorithms” etc. I found Coursera. I felt like a kid in a candy store again – there was so much useful content on data science and python, and the best part? They were free!
Nowadays there are tons of online learning resources: Coursera, Udemy, edX, LinkedIn Learning just to name a few, and of course Stanford online learning is a a great resource as well. Beyond that, lots of e-books are available online like this one, as well as tons of papers like this one.
Just remember, it’s about actually learning the material by doing the lessons and practicing it repeatedly. Work on the open source datasets, do Kaggle competitions, study other people’s code on Github. There are no short-cuts to obtaining a skill. You have to internalize it. Once you begin to internalize it, things will start to make sense across seeming disparate things and ideas will start to connect with each other. Then you know you’ve got the basics down, and welcome to the never ending quest to achieve mastery…
Step 5. Mentoring and Pivoting
“Monkey see, Monkey do” doesn’t quite work in the case of data science. The reason is analytical problem solving, critical reasoning, tradeoff assessment is all happening inside the brain. When you look at someone’s python code, it always looks so elegant and the results are pristine; but have you ever thought about how hard they had to work/re-work to get to that point?
For better or worse, all of that thought process is hardly ever transcribed, and often is the case those invisible intelligence is what represents the essence of a data scientist. So how can we overcome this “lost in translation” problem? Find a mentor. (If you are interested, come talk to us @tresl coaching)
It’s invaluable to be able to be a fly on the wall and hear an expert speak from experience how to approach a problem; what’s their thought process, what assumptions did they make, what mistakes did they make along the way, and why they chose X vs. Y. The quickest path to enlightenment I believe is to find a mentor who can show you the way.
Try to ask all the questions to the mentor and really study their response and thought process. Not only will it help you avoid missteps along the way, it’ll help you generalize things and won’t miss the forest for the trees. I was fortunate that although I didn’t know of any data scientists at the time, I had many friends who were working in top tech companies as engineers and were familiar with Machine Learning. I could always tap them for some advice or bounce ideas off of, and from their responses I could cobble together enough bits of information to find my path.
Well, even after all of the effort above, you might still be on the outside looking in. Hey, that’s me after 6 months. When I put 2 and 2 together, I figured that perhaps the industry gap is still big. I just don’t have enough relevant experience in the areas top tech companies are looking for in a data scientist. They like me enough, but the uncertainty of how I’d perform without seeing evidence in that context was still too risky. That’s why you need to pivot.
Just like rock climbing, if you only need to shift 1 limb while the other 3 are staying put, you are in a very stable position and should be able to make the transition easily. If you had to move 2 limbs, then you are probably in a very precarious state. If you only have a grip on 1 arm, well, it’s truly a leap of faith at that point.
Let’s imagine that your family (friends), your surrounding (city, culture), your industry, and your job responsibilities are the 4 anchor points. Then, if you moved across the coast with your family to a job in the same industry and had very similar responsibilities, you’d most likely be able to make the transition. For me, I had to change industry and job responsibility, and I realized that I can continue on the long-shot, or pivot to something that would get me closer. I chose the latter and joined a digital marketing agency that promised lots of opportunities to work with social data. Ultimately, that experience helped propel me into my current position at LinkedIn.
As you contemplate your own career transformation, ask yourself: how many anchors am I letting go of? If the answer is more than 1, then be open to a pivot in your journey. Remember, life is not a straight line. Every journey takes on a different set of twists and turns; as long as you are making progress towards the goal, enjoy the scenery along the way.
So, there you have it – my own career transformation summarized in 5 actionable steps. I’ve tried to distill my thoughts as succinctly as possible, and hope that by reading this post, I can help accelerate many more career transformations into data science and other areas. I leave you with one final quote, and happy learning!
Intellectual growth should commence at birth and cease only at death. -Albert Einstein
John Chao is Senior Manager, Analytics and Data Science at LinkedIN