✂️Breaking into Data Science — Getting the DS Offer & Next Steps (Ch.4)🔍
Tl:dr 💁️ ➡️ 👩🔬;
How I went from working the front desk of a hair salon to working as a data scientist without a Master’s, PhD, or quantitative college degree.
A few words before diving in…
Life comes at you fast. Since I published Part 3 (and started writing the first draft of Part 4), I’ve:
- taught at two data science programs (Springboard, DSDJ)
- consulted for a real estate start-up
- spoken at a number of conferences and career panels
- received a scholarship from the AnitaB.org for the Grace Hopper 2020 Conference
- and (spoiler alert) got an offer to work as a Data Scientist II at Livongo.
And COVID happened (and is still happening). We’re seeing a resurgence in hate & domestic terrorism, communities of color being completely ravaged by COVID, systemic racism, and an entirely insufficient administrative response to the pandemic.
I also came out as a Queer Techie to my professional network and started owning that identity, both professionally and personally.
Through these experiences I’ve grown as a data practitioner as well as a queer, asian woman in tech. More importantly what I want out of my data science career for the next 5 to 10 years has changed (even over the last year).
I initially started this series of medium posts to explain how I got my first data scientist job, intending to conclude with a happily-ever-after ending of getting the offer and settling into the role (without any major changes for a few years). And whether it’s because I took so long to finish the series (or started it with a naive perspective of what a data scientist should be), the ending instead is more of a “To Be Continued” moment as I take the next step in my career.
All this is to say that there’s life outside the budget spreadsheet, or IDE, and I’m hoping to be able to announce some exciting changes and moves over the next couple months.
Thanks for being so patient and please enjoy the last part of the mini-series (but not the last words from me!).
Part 4: Landing the Data Science offer
Young men, have confidence in those powerful and safe methods, of which we do not yet know all the secrets. And, whatever your career may be, do not let yourselves become tainted by a deprecating and barren scepticism, do not let yourselves be discouraged by the sadness of certain hours which pass over nations. Live in the serene peace of laboratories and libraries. Say to yourselves first : “What have I done for my instruction?” and, as you gradually advance, “What have I done for my country?” until the time comes when you may have the immense happiness of thinking that you have contributed in some way to the progress and to the good of humanity. But, whether our efforts are or not favoured by life, let us be able to say, when we come near the great goal, “I have done what I could.” — The Life of Pasteur (1911), Volume II, p. 228
My interview for the Data Scientist II role in Growth Marketing at Livongo was fairly typical in structure.
The interviews were as follows:
- Recruiter Phone Screen
- Hiring Manager Phone Screen
- Take-Home Case Study (involved creating a simple model)
- Onsite with two members from the Growth Marketing team, a senior member of the DS growth team, DS hiring manager, & a whiteboarding/SQL questionnaire from another senior member of the DS Growth team. Each of the interviews were 1:1.
- Follow-up phone call with DS Hiring Manager to brush up remaining questions.
I had a few other interviews in the queue including final rounds. When the verbal offer came in from Livongo, the role seemed to be the right fit at the right time. Mission-driven company, big data science org (at least bigger than the analytics/business teams I’d worked on), and a data science offering that was still evolving.
I accepted the offer approximately at the end of September, four months after completing Springboard in June 2019 and 2–3 months after starting Data Science Dream Job.
After accepting the offer for DS II at Livongo, I went on to also mentor at Springboard for a few months and Data Science Dream Job for about seven months.
It was a hectic period full of learning and productive growth. I was mentoring at two data science/analytics programs, working as a data scientist full-time, consulting part-time (while maintaining a 20 hr week gym schedule). I got to participate in datathons, speak at conferences for organizations like Lesbians Who Tech. I was helping to answer and help tons of students and aspiring data scientists via LinkedIn (enough that I started copy-pasting responses).
Part 5: Be Careful What You Wish For…
And from the moment that I saw you, I knew you was trouble,
But I disregarded detour signs,
And did not stop til you was mine.
I guess God was like, ‘Aight, fine.’
Careful what you wish for, cause you just might get it in heaps.
Try to give it back, He be like, ‘Nah, that’s yours to keep.’ — Mos Def
Although I was doing everything I thought a “real” data scientist should be doing, I was also starting to burn out. I realized I was at the end of my rope when I began to get angry and feel tired all the time doing really simple tasks.
I’d get another message on LinkedIn for help, another request for speaking, staying up later to answer messages on LinkedIn & getting headaches from the lack of sleep.
Even after dialing back on a few activities and being more strict about maintaining boundaries, especially online, I realized that I still loved parts of what I worked on but not everything. I needed to become more selective about the areas of data science and machine learning that really interested me and were worth putting in energy (and will no doubt continue to interest me for years to come).
Then COVID and quarantine hit and a lot of plans went belly-up and it’s been a struggle since to get back up and push through on some of the same goals. I realized that at the end of the day, if you’re not doing the kind of work you want to be doing, it’s not always enough to change your perspective and mindset. Grinding it out & hoping for the best isn’t always the right answer. Sometimes, as Seth Godin points out in The Dip, you’re in a Cul-de-Sac (a dead-end) as opposed to a Dip (a temporary setback that you can push through).
My time in quarantine made me realize that if I didn’t start making moves on the projects I care about and taking risks, my future self and goals would eventually be left behind.
I took the big step last month to pass on my best wishes to my peers at Livongo in their new phase as members of Teladoc without me, as I work on a new venture in real estate tech.
My goal is also to take personal time to focus on growing and developing my skills in key areas of machine learning and AI. A personal AI Fellowship if you will, in the vein of the OpenAI fellowship (whose rejection partially inspired me to come up with my own deep learning & computer vision learning sabbatical — thanks y’alls!).
All this goes to show that life comes at you fast and as the saying goes, “Man Plans, and God Laughs.”
Part 6: My Key Takeaways on Breaking into Data Science (What I did right & What I would do differently)
“Great people do things before they’re ready. They do things before they know they can do it. Doing what you’re afraid of, getting out of your comfort zone, taking risks like that- that’s what life is. You might be really good. You might find out something about yourself that’s really special and if you’re not good, who cares? You tried something. Now you know something about yourself” — Amy Poehler
Nonetheless I’m proud of myself for having broken through that first major milestone (getting a data scientist job) and if I were to go through the process again (say a year from now), here are the things I would do differently (or not).
The 5 Things I Did Right
1. Getting started exploring early
It’s really easy for time to just creep up, especially when you don’t have a clear goal in mind. Some would even say time seems to speed up as we get older. One reason why time can seem blurry is because once we get past college (or grad school), there are far fewer transformational achievements and external pressure (and when there is external pressure, it’s to conform). And when you combine key milestones being pushed back further, with many people taking up to 20–30 years to hit their peak potential, there’s a good chance that delaying your exploration means you won’t see real gains from switching to data science in your working lifetime.
All of this is to say, do the math.
If it takes a year of study, applying & interviewing, then an addition 2–3 years to get to a senior level, it’ll be at least three years from the time you commit.
Get started earlier so you can reap the benefits earlier.
2. De-risking as much as possible, when possible (& leverage Tripod of Stability)
Students & mentees have asked me whether they should quit their current jobs to study or do bootcamps, whether they should turn down jobs that aren’t a fit while interviewing for other reach roles. Most of these “burn your boat” questions depend on your individual risk tolerance and ability to recover from the consequences.
In my particular situation I was blessed with some safety nets that included a below-market rent controlled apartment (thanks SF!), a partner that had a full-time job (and rent controlled apartment) that was supportive of my career, living in the same city as my parents (who would never let me go homeless), and an emergency account I had built up through aggressive saving over the prior years. My students loans were also fully paid off, car was mostly paid off and the remaining monthly payments were incredibly low. Between all these safety nets I could have quit my job.
But I didn’t because I have a deep seated fear of debt, being homeless, and not having money when I need it. I was willing to trade time in exchange for keeping money and liquidity. One of my life meta-goals is to always keep my options open and to even build-in options where possible (for example, learning a new skill).
Ramit Sethi advocates for a form of risk-mitigation called the “Tripod of Stability”.
The idea is to keep “three of the most important aspects of your life perpetually stable. This stability gives you the confidence to take the occasional risk.”
In my case, my tripod of stability was my job/income, my core relationships (partner and family), and my routine (work outs, date nights, & family dinner nights). If I decided to quit my job I would have been incredibly stressed out and felt pressure, potentially risking exposing myself financially (no income, drawing down on my savings to pay program) and breaking continuity in my resume.
In general I wouldn’t recommend quitting your job in order to pursue a graduate degree, certificate, or bootcamp (especially if you need to take out a loan). Nowadays most bootcamps have an ISA (income sharing agreement) where they garnish a % of your income once you’ve gotten a job in data science (up to the cost of the tuition plus interest). Typically these ISA’s are a much better option than a loan from the same bootcamp in most cases and a better deal than loans for Master’s programs. ISA’s put the onus on the camp to provide career assistance in the form of resume reviews, practice interviews and referrals to prior alumni.
Instead of quitting your job I’d recommend starting your data science endeavors by taking some online courses (preferably with project based components) that are available asynchronously to determine if you have an initial interest. If you take some courses and hate it, then all you’ve lost is a few weeks or months of time and maybe $20–100. Even in that scenario you’re probably learning 1–2 skills that you can utilize in your own work. If you do really like the courses, I’d then recommend getting a hold of a dataset from Kaggle and doing an end-to-end analysis and modeling project (you could even use Kaggle or Google Colab).
3. Being fully informed & doing the research
I’ve gotten questions over the years about whether data science bootcamps (like Springboard) and job search prep programs (like Data Science Dream Job) are worth it. I’ve covered my experiences in both Part 2 of this series as well as different promo pieces for the programs and encourage you to read about my reasons in more detail.
I do want to emphasize that what works for one person might not work for another.
Some factors you might want to consider when choosing a learning medium (and provider) include:
- Whether you’re just starting the data science journey or have been at it for a while? (i.e. do you need to cover all your foundational bases or do you just need to deep-dive on specific topics)
- Whether you’re someone that needs structure and hand-holding or someone that prefers a more hands-off self-initiated approach? (i.e. do you need a step-by-sep program with clearly defined deliverables or are you okay with a curated list of resources)
- Whether you’re someone that needs a consistent, recurring presence or prefers to reach out when you’re stuck? (i.e. do you need a mentor or a support community or both)
- Whether your financial situation requires you to continue working full-time or you’re able to take significant time-off? (i.e. does the material need to be asynchronously delivered or can you do live learning)
- Whether you have other needs besides learning the skills? (i.e. Are you hoping to have a portfolio or projects?)
- Can you get into the program? (i.e. Is there a GPA or prior experience requirement?)
When considering the specific bootcamp (and whether I went the bootcamp route versus getting a degree) I considered and researched all the questions posted above and ended up with the solution that fit me best.
I’d encourage you to also do some research and scenario planning and take everyone’s endorsements with a grain of salt.
Bootcamps can provide a valuable environment for candidates to strategically bridge areas of weakness, create professional relationships & benefit from group learning (but it doesn’t mean they’re right for everyone).
4. Do whatever it takes to succeed
One of my favorite books in the world is “So Good They Can’t Ignore You” by Cal Newport. I first picked up his book in 2013 as part of a personal mission to climb out of the hair salon and into tech and I’ve re-read it many times over the years. A fantastic summary of the book’s central points can be found here but the most salient theme is about how experts leverage a “craftsman mindset” and work on building “career capital” by deliberately and intentionally developing and growing “rare and valuable skills”.
At the start of the Springboard Data Science program I was struggling with timelines and getting work done. I felt like I was putting in hours here and there and squeezing in as much time as possible and still unable to make real progress.
One night I caught up with my friend Veena, who is equally (if not more) growth oriented and a Renaissance woman that works as an engineering director in the day and spends nights singing and playing (as well as composing) music. She is an incredibly multi-talented individual (that also gets business done) and I was really interested in how she was able to be so productive with the same number of hours in a day. I was shocked when I found out that in order to get her creative hours in, she was waking up at 4:30am and going to bed at 8pm (I was struggling to even get up at 9:30 am!).
I realized the answer wasn’t to continue shaving 10–15 mins here and there but to radically redesign my schedule around my goals (or as Cal Newport calls it, creating a “fixed-schedule”) and ruthlessly prioritize everything around those goals. If it didn’t contribute directly to my goals of kicking-ass in the Springboard Data Science track while maintaining my relationship with my family and partners and staying shredded in the gym, it didn’t make the cut. That willingness to make big changes was 8 months of no friends, no parties, no going out, no shopping (to maintain financial liquidity) and being a code monkey.
And guess what? I accomplished my goal and that accomplishment opened so many doors professionally, from getting a main gig as a data scientist, side gigs as a mentor, and speaking engagements and scholarships.
Accomplishing transformational goals isn’t fun and they require you to change as a person. At the end of the day, I’d like to think that we’re aiming to be the people we never thought we could be but had always wished, and sometimes that requires equivalently big changes.
In the words of Mr.Carter,
“People look at you strange saying you changed
Like you worked that hard to stay the same
Like you’re doing all this for a reason
And what happens most of the time.. people change
5. Going all the way
Everyone has probably seen Angela Duckworth’s TED Talk on Grit and her formulas by this point (and if not, it’s worth watching the TED Talk and reading her book). One of her main points is that grit — a combination of passion and perseverance for a singularly important goal — is the hallmark of high achievers in every domain. Another one of her quotes is “Talent counts once, effort counts twice”.
I was never considered a particularly smart individual and certainly not an academic achiever in my early years (writing and journalism scholarships aside). I didn’t go to an Ivy League or graduate as valedictorian. And I wasn’t completing a prestigious graduate program. I could have let these doubts about my ability to get a foot in the door prevent me from succeeding.
What helped me sustain the journey was learning about and developing a growth mindset with the help of mentors-at-a-distance like Carol Dweck . I constantly reminded myself that data science is a collection of skills and skills can be learned. I knew the biggest meaningful differences between myself and other candidates (that might have gotten an earlier start in data science) was time and intention.
So as much as possible, I incorporated learning throughout the entire day.
For example:
- I’d spend the first 3 hours of the day going through coding tutorials and videos;
- I’d print out notes or powerpoint presentations and study on my bus commute;
- I’d use lunch to do more tutorials and coding, as well as the afternoon rush hour (the bonus of hanging out at the office after work ended was healthy snacks and never ending good coffee from Philz);
- I’d listen to videos and lectures on the stairmaster at the gym;
- Read some more on the bus ride, rounding out about 2–3 hours of commuting.
- And then another 2–3 hours of working on projects.
Using the dead time on the bus commutes alone I finished reading “Practical Statistics for Data Scientists” in 1.5 months.
The 5 Things I Would Do Differently
1. Be less picky about the datasets & projects
When you’re starting out and have no projects in your portfolio, it doesn’t make sense to try to find the perfect dataset and the most unique project under the sun.
When you’re first learning about the different algorithms and supervised learning tasks it’s actually better to choose well-documented and bench marked datasets that are publicly available on Kaggle.
There’s a number of reasons for this:
1. Well-documented datasets have tons of sample projects and code that you can compare and contrast;
2. Because others have already worked through the datasets, it also means they’ve done the initial discovery and EDA and discovered some of the major data quality issues (as well as solutions);
3. Modeling is actually the easy part of the data science workflow. Data cleaning, feature engineering, model interpretability and prediction deployment are the hard parts. Working on the easier parts of the process helps build momentum early on in your learning for when you tackle the hard parts.
Instead of trying to find the perfect dataset (or even generating your own through scraping) focus on improving your EDA, modeling, and presentation skills. My Springboard completion got pushed out by two months because I was so busy trying to find interesting datasets and then setting up the ETL’s that I ended up scrambling on the other parts of the projects.
2. Understand the whole data science workflow, instead of rat-holing on specific areas
Personally I despise all the tool talk on LinkedIn. R vs Python, AWS vs GCP, etc. At the end of the day, the majority of us aren’t and will never be in positions where we determine the tech stack for a company. Most of the time you’re adapting to the tech stack that already exists and adding or subtracting personal peripherals (Jupyter, JetBrains, VS Code, who cares, it’s your choice).
So why do we have so many people engaging in battles over tools or point-solutions?
Part of the answer is about the social clout and clicks i.e. there is nothing that gets people more riled up and “engaged” than lambasting another person’s favorite programming language. (And “5 Reasons Why ‘R’ Sucks” is easier to fit in a subject line than the more nuanced and reasonable title “The Trade-Offs and Considerations in Supporting R & R Studio for Non-Academic Analysis in Production Teams”). It’s also easier to copy-paste the features list from the Conda site and rewrite them and add useless commentary than go into the different trade-offs and use cases.
Another part of the reason, I would hazard, is because some practitioners don’t understand enough about how data science and machine learning work in production and oversell the value of machine learning models. At the end of the day if your model isn’t integrated into the business and usable, it’s just another toy project that didn’t bring your business or organization any value. And if you don’t understand the entire workflow (from data to model to prediction to evaluation to serving to monitoring) it makes it harder to figure out what the solution should be for your problem.
When I was working on my NLP project predicting kickstarter campaigns for example, I struggled understanding NLP techniques for feature engineering (tokenization, n-grams, etc) versus NLP tasks (like measuring document similarity, etc). My understanding of the entire modeling pipeline was poor and I ended up spending more time reading up on NLP tasks as opposed to applying NLP techniques for feature engineering and getting creative with the feature engineering and model interpretability (instead I ended up slapping some basic classification algo’s on top of the entire mess).
By now you’ll notice the general theme is that the more you understand about what the process should look like, the less time you’ll spend on the low-value parts of the pipeline.
3. Spend more time on the basics of interviews, as opposed to the fancier, edge-case questions
Because interviews are time-bounded, there’s only so much they can ask you to do (especially now with remote interviewing). I wish I’d spent more time on reviewing the basics (experimentation, EDA, model evaluation & deployment, SQL, leveraging standard python data structures like collections objects) and less on edge-case questions, especially for business data scientist roles. For business data scientists it’s less forgiving if you get a stats or python question wrong than a question about data processing for computer vision models.
If I’d spent more time on going through all the python questions on hackerrank and the python/SQL questions on codewars (and doing the questions on a whiteboard), I probably would’ve had a better chance nailing offers for some of the dream companies out there.
4. Don’t ignore red flags in interviews
If you’re seeing red flags in the interviews about the specific role you’re interviewing for and the team (and the data science maturity of the company), it’s only going to get worse. It could eventually get better but trust that it’s usually a lot worse below the surface.
People who are happy with their jobs and really excited about the company will let you know upfront. When people hiss between their teeth or go “Well, you know, things are always changing” as an answer to a rather direct question, swipe left.
5. Marathon, not a sprint
Learning and growing as a master of your craft is a life-long quest and journey. You need to continue developing your skill set and can’t stop just because you get an offer. I regret getting lazy after getting my job and focusing so much on signaling (conferences, speaking, etc) and dropping the “craftsman mindset”.
Thanks for reading this far!
I finish up the series in my next post by answering the most common questions I get in-person and on LinkedIn.
Interested in following along with my Data Science journey? One of the easiest ways is by connecting through LinkedIn or GitHub. Feel free to send me an invite and let me know what you thought about my series!
And if you really love what I wrote, consider buying me a coffee at www.buymeacoffee.com/mmbazel & keep me writing! ☕
If you’re curious about the series or want to jump between parts, feel free to check out the rest of the series:
● [You Are Here✔️ ➡️] Chapter 4: The End is Just the Beginning + FAQs (aka “The Offer and Introspection”) — This section covers my offer and summarizes the high and low points of journey to “becoming” a data scientist and next steps. I’ve also expanded this chapter to include the top questions I get about becoming a data scientist. Key Takeaway: The End is just the beginning! 🎉
