My journey from academia to data science

Well, I did it. I'm gainfully employed as a data scientist. Every person that makes this transition has a unique story to tell thanks to that individual's unique set of skills and experiences and the particular course of action taken. I can say this because I listened to a lot of data scientists tell their stories over coffee (or wine). Many that I spoke with also came from academia, but not all, and they would say they were lucky...but I think it boils down to being open to possibilities and talking with (and learning from) others. For what it's worth, here's my story. Consider it another data point.

I was always interested in the sciences and eventually achieved a PhD in Neurobiology. My day to day involved designing experiments and making sure I had ingredients for recipes I needed to make for my experiments to work and scheduling time on shared equipment. (Thinking about grad school or in a program now? I bet you didn't count on being a logistics expert.) I knew for the most part what statistical tests would be appropriate for analyzing those results (when the experiment worked). I was getting my hands dirty doing experiments at the bench (aka wet lab), but the fanciest tools I used for analysis were Excel and proprietary software called GraphPad Prism.

In the summer of 2015 while I was a senior fellow, I took a seminar offered at UW called Dependable Strengths that helped postdocs and grad students identify skills that they enjoyed using in academia and how to market that for jobs outside of academia. Let me plead passionately here to all you grad students considering careers outside academia to please visit your institution's career center. I wish I did this sooner, when I was still a grad student when these resources would have been FREE. It was an excellent motivator to get a resume put together (no, recruiters will not sift through a CV) and feel confident that I could start informational interviews. If this last part is a strange statement (what's so hard about asking someone to sit down with you and tell you about their job?), I was constantly convincing myself that I wasn't ready for this step, that it wasn't the right time, that I didn't have anything useful to provide in return, that I was bothering the other person.

In the fall of 2016, I attended a panel discussion of UW alums working as data scientists which was in preparation for a computer science/data science career fair. I started applying for analyst positions and did not progress past take-home technical assessments. At this point I was tinkering on and off with learning Python in Codecademy. Even though Python was new to me, programming was not. I had taken a few programming courses during undergrad way back when as part of my Cognitive Science degree, but bad experiences resulted in the wish to never write another line of code. The irony.

My job search slowed as I read more data scientist and data analyst job descriptions and saw that SQL, Python, problem solving, creating visualizations, presenting to C-level executives were skills in common. I had some of those skills, but not all of them. This was a solvable problem though: I problem-solved on a day to day, so check mark on that, so I just needed to pick up some SQL and Python right?!

Sort of. I completed tutorials and worked on exercises, but I felt lost in the sense that I wasn't accomplishing anything worthy of putting on my resume. I didn't know about Kaggle datasets or what machine learning was and it's difficult to research something if you are unaware of it's existence. A friend encouraged me to attend a ChickTech event which included introductions to bootcamps in the Seattle area. I didn't want to pursue a second postdoc in bioinformatics or computational biology or a start a master's program in data science, so a bootcamp offered a way for me to get up to speed and into the job market. This also meant leaving my postdoc.

I researched bootcamp options: Galvanize, Metis, General Assembly, Insight Data Science, and The Data Incubator. If you have an advanced or professional degree, Insight and Data Incubator are options to look into as 7-week, intensive fellowships (potentially free if you are accepted), however (spoiler alert) as I did not want to relocate, I ultimately considered other options. Note that Insight currently has a Seattle campus, but this wasn’t the case when I was looking. I attended open houses to ask questions about the curriculum, types of support provided during and after the program, and talked to admissions officers to learn about their expectations of prospective students. Thanks to these conversations, I was given resources to study from and in the evenings after work for the next few months, I dove into statistics, programming, linear algebra, and brainstorming project ideas (topics covered on the admissions exams). One thing to note is that I narrowed down my choices of bootcamps to Metis and The Data Incubator before applying because the admissions process was fairly unique to each bootcamp; I personally could not invest the time and energy to apply to every program. Unlike applying to undergrad or grad school, I couldn’t write an essay and make adjustments to suit each program, rather, I had to schedule and take a suite of tests specific for each program. The tests and the acute stress that came along with it was a lot to deal with, however, unlike applying undergrad or grad school, I found out whether or not I advanced to the next round in less than a week. Metis is a 12-week program, similar to General Assembly and Galvanize, and has the distinction of being ACCET-accredited. I was interested in Data Incubator because they were opening a Seattle campus (and there was the possibility of receiving free training).

With upcoming application deadlines, I applied to these programs, each involving a few rounds and starting with an online programming assignment. Passing that, I took a timed take-home assignment, mostly multiple choice questions of in the above mentioned topics. Data Incubator also required a video recording of the project pitch and at least 2 (3?) recommendations from your professional history. Lastly was a Skype video interview with a data scientist. Please contact the bootcamps of your interest directly for the most up to date information regarding admissions :) When I was admitted, I chose Metis mainly because Data Incubator would not have their Seattle location open in time. I gave about 3 months notice (a blink of an eye in academia) and wrapped up as best I could in lab. I wanted to allow myself some time to breathe and deal with the fact that I was moving away from a field I had studied and trained in for the last 15 years.

Then it was down to business with the prework. Yes, you heard me! You have homework to do before you even step foot onto campus! I was told to allocate 60 hours and as someone who did not come in with the strongest programming background, that was a good estimate. Metis also strongly advised incoming students to complete prework and the bootcamp using a Macbook Pro (I believe this is common across bootcamps after I had the chance of meeting other bootcamp fellows). As much as I wanted to avoid adding to the cost of my professional pivot, it was an easy decision to go find (a refurbished) one. It was tempting to use my trusty (PC) laptop that I already owned, but I was told that the curriculum and instructions for installation days (e.g. ssh to AWS, Docker setup) were all written with iOS in mind. I did not want to add this type of troubleshooting/distraction/multitasking to my main task of learning data science. I stand by that decision and am glad I made it. On the day I set out to find my prize, I was lucky (there’s that pesky word again) that a local store had a 13” 2015 model come in that morning. I experienced a steep learning curve when making the transition to iOS, but I can more or less switch back and forth now.

During bootcamp, I full-mindedly and whole-heartedly focused on data science. I was in the classroom from 9:30 am until at least 5:30 pm for lecture and classwork, most of the time longer because of practicing data scientists that came to give talks or to work on HW or projects. Once or twice a week, I would attend Meetups (e.g. PyLadies, PuPPy, Ballard Tech HH) or other networking event (e.g. New Tech Seattle, WiB). One week I attended three social events and discovered that was too many for me, which, I know, sounds crazy on the surface. What’s so bad about meeting cool people and eating free food and drinking free booze at the same time? If you are an introvert, this probably qualifies as a fresh circle of hell, but try it out with a beverage in hand and a “hello”. I went to a few different Meetups to find my tribe and investing in that effort is worth it. If I can relate this back to grad school, you remember what it was like to attend the social events after the full days of interviews as a prospective student? Being fun and smart and _on_? It was a similar situation. If I had time and energy when I got home, I would work on a project, homework, look ahead to see what we were going to cover the next day, go through my notes on a concept or technique that needed going over. I’d go to bed between 11 pm and 1 am (later if it was crunch time during a project deliverable week). Rinse and repeat for 12 weeks and that was my bootcamp experience in the smallest of nutshells.

Career Day capped off the bootcamp portion of the Metis experience, was exhilarating, and signaled the beginning of the journey to come. The job search required both focus and flexibility. I (and probably you) have heard that job searching is like a full-time job and it seems accurate in the sense that can you spend 6, 8, or more hours doing things related to finding a job. However I found it hard to dedicate big blocks of time to applying on company job sites. I continued attending Meetups and by now I found groups that I connected with, so knowing that I would see friendly faces at the end of the day was motivating. Through a mix of cold outreach and warm intros, I also continued reaching out to local data scientists on LinkedIn to learn what data science meant at that company and to ask questions relating to day to day operations, the size and personality of the team, what the backgrounds of recently hired team members were, etc. These were wonderful opportunities, often led to being connected to other individuals and directed to other resources, and in some cases helped me decide if I should continue investing energy to apply to a position. I was job hunting as we were headed into holiday season (Thanksgiving, Christmas, New Year’s) and staying connected to people helped me keep the momentum going when interviews slowed down. I wasn’t kidding about the coffee and wine.

I had a whole list of things I wanted to get done and found that I was checking few items off at the end of the day. For me, writing down a to-do list wasn’t enough. It was too easy to spend more time on things I wanted to work on. What ended working for me was using my Google calendar to visualize and schedule everything. I scheduled blocks of time for job applications, reworking my cover letter, Meetups, studying algos, doing coding exercises, interviews (obviously), watching youtube videos, reading, blogging. Once I saw what I wanted to accomplish that day and how much time I was willing to spend on each task, I filled in the rest of the time with when I’d take lunch, go running, and perhaps most importantly, I scheduled “Wild Card” time periods for things that might give me a headstart on the next day or an opportunity to circle back to something I worked on earlier. This is where flexibility and fudge factor time fits in. Some company ATS being particularly ornery and worried about not having enough time for reading important articles? Wild Card time saves the day and I don’t end up sacrificing time for things like self-care.

Fridays were for check-ins at Metis (cheers to career support!) and spent specifically on networking, mainly on LinkedIn. During the week, any people I needed to get in contact with or follow up regarding informational meetings, job leads, etc, would go on my Google calendar in the description field. When Friday rolled around, I worked my way through the list. After filling in my calendar for the upcoming Monday, I called it a day. And I took the weekend OFF. If you are anything like me, I felt guilty for doing this. In academia, there was always some journal article I should be reading, an experiment I could be setting up (gotta take advantage of free weekend/holiday parking!). It took a few tries for me to work out My System, but figuring out how to keep myself focused and efficient made my day run smoothly and removed excuses I could come up with about feeling guilty.

Interviews! There’s no other way to say this other than, data scientist interviews are challenging to prepare for. Even if a company provides a study sheet, the questions asked can range a variety of topics: talking through case studies, technical questions about SQL functions, pros and cons of various algos, issues to be aware of when fitting a model, whiteboarding in Python and SQL, general experimental design, etc. I personally wasn’t asked probability or too many statistics questions, but they are fair game depending on the type of data scientist position or interviewer. I also hate to say that sometimes you might not know what you are up against until you completely bomb an interview. It happened to me and it might happen to you. It did not feel good, but if you are anything like me, you make sure you don’t trip up on a similar question the next time around. (Because you’re going to make flashcards for those, right? Ok, good.)

But you don’t just get an interview just because you applied for a position. I sent out about 80 to 90 applications in a 5 month period and getting that initial email asking to schedule a phone call was so exciting! For each company, it was common to have an initial half hour screen with the recruiter who posted the position asking behavioral questions (my background and experience and what I was looking for, etc). I’d hear back within a few days whether or not I would move forward in the pipeline. If it was a “no”, I would ask for feedback, but normally wouldn’t get anything more than “there were many other qualified candidates, many who had the experience we are looking for”. However sometimes hiring needs change and the position you originally applied for gets put on hold. Sometimes the posting was meant for an internal hire. There are lots of factors at play, so don't feel bad and don't automatically blame yourself. For a “yes”, the next phone round was either with the hiring manager or someone technical (sometimes one and the same) who would ask job related questions or anything that I mentioned in the paragraph above. Sometimes the next round was a screen share with an individual contributor level person (potentially your future teammate) who would ask me to code and ask other technical questions. Sometimes both were done….it really depends on the company/department/team. At this point, if I was moving forward, it was time for the in-person. This is always a marathon session. My in-person rounds happened to be contain more behavioral questions, questions that helped them learn how I handled difficult situations, learning new tools/skills, though there was also a good number of walking through case studies. As stressful as this situation is, it’s an opportunity for both parties (you and your interviewers/future colleagues) to see if being added to the team would be a good personality and cultural fit. If you think about it, you spend about a quarter of your time with these people (more if you work more than 40 hours a week and factor in any socializing events), so I was ready for each interviewer when s/he asked, “Do you have any questions for me?”. I would ask my questions relating to projects and tools, and I always asked at least one question regarding work culture, depending on how much time I had left. My advice here (if you were looking for any) is that you worked hard to get to this point, so include time in your preparations to learn how you will fit in and how the organization treats their employees.

Maybe I should have included a tldr, ah well. I started this post (Feb 13!) just before I started the new job and and at this point, I'm just over 5 months in! I'm learning so much in regard to subject matter (insurance is so complicated), modeling practices, and new tools...new to me (R), but also not necessarily the newest technology (SAS, whoop). Happy searching and good luck!