Back to Home
Click to summarize. It doesn't work on every page... some are short enough!

⏳ Questions Students Ask

Outside of my normal 9-to-5 job, I also teach a class at my alma mater of the University of Washington. Having taught in the UW iSchool for a few years, I often get similar questions from my students in the classroom. I've captured some of those questions below and expanded on them more than I normally do in class.

My class tends towards people who are early or mid career and are attempting to either pivot into the field of data science, or become more skilled/literate with data analysis and use that to level up there already extant subject-matter expertise.

🙋‍♀️ What do companies look for when hiring?

Ultimately it's hard to say, but it is very much dependent on the size of the company, the macroeconomic landscape, and other factors outside your control. In my experience, however, demonstrating you can learn quickly and that you have ~60-80% of the skills in the job description are the most important attributes.

When trying to get into my first job out of grad school I did a very niche project for folks and that demonstrated that--while I didn't have deep expertise in the domain--I could quickly become conversant in new ideas that I would be introduced to. I have mostly done hiring at smaller organizations, so it tends to be higher-touch experience than at a large tech company, and you can use that to your advantage.

In scenarios like this, you can do focused research on what the hiring manager is looking for, and try to position yourself as a solution to the problems they know they have, but also the ones they may not yet be aware they have.

💼 What do you recommend for building your first data portfolio pieces?

In general, there are a handful of core things I look for in someone's portfolio, either when I am exploring it as a hobbyist or for hiring purposes:

End-to-end work that demonstrates varied data science skills

An example of this might be going from research question/business case to implementation in code/infrastructure. It could also be analytical work on an already-extant dataset to otherwise summarize or communicate information, which may or may not include visualization. Sometimes these projects are essentially self-documenting, like if you put together a Jupyter notebook with your data analysis in it and good commenting. Others, like building a data processing pipeline, may need more context and are not at all naturally salient to the person approaching your work.

A project of interest or curiosity which hasn't been overdone

This shows you're a creative thinker and can approach problems from different vantage points, plus the subject matter is a differentiator unto itself. For example, instead of doing the typical Titanic random forest or iris k-means, do a data analysis project with the Great British Bake-Off or RuPaul's Drag Race. I know lots of people are excited about working on these, but there's diminishing returns in terms of the ability to find something new in these datasets.

Levels of information accessibilitythat show communication & collaboration skills

This is a major differentiator. Maybe you summarize everything into a nice blog post and walk through your analytical process, but then you share your source on GitHub in a well-documented manner. Showing how you can actually complete the task is one thing, but being reflective and having the capacity to share your choices or decision points along the way are equally important. Especially in data science, it's not just about the code.

The best candidates have at least 2 examples like this on their personal site, but possibly more.

Typically I weight an ongoing programming or data analysis project higher than I would a one-off, because it demonstrates consistent engagement with an idea over time, but I don't discount anybody for having tons of small projects if they are all complete.

🗺️ How should I show off what I produce?

Build your portfolio on a web platform you control, one which is (mostly) not subject to the whims of a large-scale tech company with different incentives than you. There are many different ways to do this at different levels of affordability.

Here are a couple of recommendations that might be helpful to you:

🔎 How do you recommend tailoring your job search for the first time, especially coming from a discipline that is not focused on technology?

Augment your supplemental materials - While they won’t explicitly get you a job, a portfolio, reel, or showcase of your work is invaluable if you’re going to attempt to transition into the space from a field you don’t have professional experience in.

If you can do an end-to-end project which captures use of your skills in some way, that will both level you up re: domain knowledge (for example, something my peers in school did was salad taxonomy -> ML classifier; based on salad metadata, assign category -> explanation of results) but also act as an artifact that supports your candidacy. It’ll give you something to talk about.

Be realistic about the process - It will absolutely take time to get you your first job, and you shouldn't expect to convice others that you are a new kind of information professional just because you have a diploma now.

If you're like me when I was in school, you’re trying to switch careers into a new domain and also learn the technical aspects of the domain and also learn the domain's new professional communication tactics, rhythms, and customs. Each of these things separately could be a full time job. It’s really helpful to have conversations with other people who are doing similar work, support each other, and keep in touch about learning and growing. Honest, engaged human connection on the job search is powerful and like rocket fuel in some situations, and is what real networking is. Being kind and vulnerable with others is a form of being kind to yourself.

Learn to notice - Part of the work you have to do right now is understanding how filtering works to determine good jobs for you vs. bad ones.

At this point, avoid the term “senior” and optimize for “junior” or jobs without a qualification at the beginning (senior may have management duties and would be less focused on technical expertise, but would still require domain knowledge, so you probably wouldn’t be well-suited). Read the descriptions closely and understand what what the firm is looking for as best you can—keep track of words and noun phrases that reoccur and try to do research into them if they’re unfamiliar. Look for trends in jobs that you consider interesting--what are the commonalities? How might you find more jobs like this?

Build in reps as fast as you can - Try to get your first conversation with a hiring manager or recruiter as early in this process as possible.

Ask for feedback when you can (some organizations of a certain size don’t give feedback). Some recruiters, being time poor, will end a conversation with you early if you’re not a good fit and tell you why, which is definitely to your advantage as rejections go. All the information you can gather here is a scientific experiment—what are the different things you can say to represent yourself or convey your experience to get a step further through the job-seeking process? Build in psychological distance so you can take care of yourself emotionally. Half the battle is positioning; the other half is doing the work. You will fail, but you want to fail fast and often so you can learn faster & reposition.

Job hunting is like writing--just sit in the chair and do it - Block off a fixed window of time on a schedule that works for you to look for and apply to jobs.

This could be a daily or weekly process, but try not to do it less than that. Use affinity organizations (communities based around software you use in this domain, professional groups, etc.), LinkedIn, and job boards to find jobs that look interesting. Consider trying the following:

📕 How do I find the best resources to continue learning or professional development outside the classroom?

While in each category below there are books you can buy, I highly recommend open source curricula and courses for supplementing your education. Especially with the advent of LLMs, which are essentially personal tutors you can ask any question to in any level of detail, you can use the subject-matter expert's outline of the area as a conversation starter with your preferred chatbot. You're likely already paying enough in tuition; no need to spend more money.

When I worked at the University of North Carolina helping faculty members with course design, one professor did something I liked a lot. She presented the fundamental elements of each unit to students as required readings, and then paired them with a section called "If You Want to Know More" which outlined specific topics of interest. In this section, there were materials designed to augment the narrative threads that often came up in class or in the other reading materials. You can treat this outline below as the same thing!

What I would suggest as a process is something like the following:

  1. Pick up one of these books and start to go through it
  2. Encounter ideas you find confusing, unclear, or otherwise don't have a reference for
  3. Ask your favorite LLM chatbot to explain the same concept in a different way
  4. Build a mental map of the concept space
  5. Venture into different levels of detail surrounding the topic with your new tutor

Programming, Computational Thinking, and Systems Thinking

Across the board I highly recommend books by Allen B. Downey. His materials were a major force multiplier in my ability to learn and understand programming, after coming from a non-STEM background.

Data Visualization, Narrative, and Communication

There's no avoiding the work of the greatest. Anything by Edward Tufte is a great primer on visualization, but I think it's best to start at the start with The Visual Display of Quantitative Information.