We know each other all too well. The weekly meetups. Late nights spent on competitions. MOOC forums, and of course the “MLearners-Anonymous” version control support group. You live an unbalanced lifestyle because you feed a nocturnal addition: data science. I was in the same boat once, expending egregious amounts of would-be family time honing the skill. Then I took a leap of faith, and transitioned into a role where I could do what I love full-time.
You can too.
Here are a few tips on how I landed offers at Fortune 100 and startups alike for the following roles: Lead Data Scientist, Data and Applied Scientist, Machine Learning Engineer, and Principal NLP Engineer.
People I’ve met along the way, both self learned and PhD’s alike, mistakenly felt the need to have placed Top-3 in a Kaggle / TopCoder / Innocentive / Grand-Challenge / DrivenData competition; or to had built out a marketable product before truly becoming hireable. Hogwash. If every practitioner had Marios Michailidis’s and Jeremy Howard’s amazing skill-set, I’d be even MORE worried about governments seducing talent over to the dark side. Being the best is awesome but not required. You just need to be good, curious, and a perpetual learner. Competing, struggling to come up with solutions, procuring and wrangling your own data sets, and working in teams are all efficient ways to expedite that.
My interviews at larger companies involved coding, where smaller companies seemed to prefer technical dialog. Be ready to talk for up to 30 minutes about a ML project you took part of and whiteboard common Spark and Pandas jiu-jitsu. Can you walk through the entire DS process from idea to production? Can you speak on cross validation, imputation techniques, NNets, and GBDTs? How many model evaluation metrics can you code from scratch? How do you interpret feature importance? These are all good skills to have.
Curiously, none of my interviews involved demonstrating mathematical knowledge of calculus nor linear algebra, however almost all of them had practical statistical questions: What do the coefficients of LogReg mean? What is standard error? When would you use a Poisson distribution? Given a random sample, how would you parametrize a distribution, etc. One interview even asked a question on Bayesian optimization (I didn’t do well there).
Standing out goes both ways. Don’t just jump on the first gig you’re offered! Consider all things that make said company stand out. How much and how are they compensating you? Are titles important to you, i.e. Lead Data Scientist vs ML Engineer? How far is your commute? How much recon on the corporate culture have you done?
So many free resources exist with tips on technical resume writing. Dedicate a week to studying and implementing them. Use grammar check and then have 2–3 recruiter friends and 2–3 technical friends one-over your work. Personalize your resume per position, and always include a thoughtful cover letter if that’s an option. If not, include it anyway as the first page of your PDF.
Before mucking out my CV, my response rate was around 1:14. After a polish it out, I was batting around 1:4. The same skills, just different text and arrangement. For starters, be succinct and try to keep it 300–600 words max. If you do a good job on your resume, you’ll have a chance to talk more during your phone screen and later interviews. List your titles and measurable relevant accomplishments. If you don’t have DS work experience, drop some of your projects and public speaking engagements to show SME.
Your network is your net-worth. Companies absolutely love hiring people their employees vouch for. Spend time building your network, online and offline, while taking MOOCs, attending meetups, and competing. If there’s an employee present from of one of the companies you’re targeting, introduce yourself then DS geek-out. A casual way to close a fun night of discussion is ask them if it’d be alright to connect with them professionally on LinkedIn. Let the sauce simmer a little, then follow-up with a well crafted message asking them what they think about the department / position you’re gunning after. Then hit them with the resume and referral request. Foot in door phenomenon, friends.
For any offer you take up, one should consider four things:
- Do you enjoy doing what you’re doing?
- Are you, your coworkers, and direct report competent in the field?
- Are you positively impacting people?
- Are you getting compensated competitively?
We have a beautiful field and industry (except a certain toxic sub-reddit) and practitioners of data science have a solid shot of striking all four. If you’d like to talk data science or want another set of eyes on your resume, tweet at me and I’ll share free personalized suggestions, time permitting.
Two last pieces of advice. Don’t ever get discouraged. I applied to almost 60 jobs spanning 56 companies (shameless, I know!), but only got interviews at eight of them. Most companies will just ghost you and not even do the courtesy of a rejection notice. Keep your head up and don’t take it personal. Internal politics, budgets or lack thereof, inactive projects, time mismanagement, and many other factors are likely at play. It’s not *always* because someone else swiped your spot. Keep trying and you’ll eventually land something.
Second, be vigilant about guarding your privacy. Amass as much info about the company posting the position as you can. There are a lot of firms with simple 3–4 page websites that just suck up resume and personal data. Don’t fall for them no matter how sweet their tune sounds, and never share your SSN until after you have a penned offer from a company you trust.