Lessons Learned as PhD Student

For those of you who follow me on TikTok , you'll know that most of my posts from the past two years have been about my trials and tribulations working on my PhD. Well, in less than 10 days I will finally be defending my dissertation at the University of Maryland Baltimore County. Before I defend, I wanted to share my Top 5 Lessons Learned as a PhD Student. If you're thinking about applying for a PhD - especially in computer science (or adjacent) fields - these might pertain to you. However, I think they're relevant for most PhD experiences.

@mybuddyskynet

Thanks to everyone for the ❤️ and the encouraging 💬! I'm not out of the forest yet. But it's a big milestone. 😁 #phdlife #AI #deeplearning #coding #datascience #computervision #machinelearning #python #generativeai #phd

♬ Blinding Lights - Instrumental - The Weeknd

Lesson 5 - Your Relationship with your Advisor is Everything

For over a decade, my "day job" has been as a consultant. I work for a very large consulting firm and fancy myself as someone on the technical R&D side of things. But regardless of whether I was functional or strategic (i.e. technical or business dominant), you learn that in a big corporation the skills of working for many masters. Of course that's being a bit pejorative. But, the idea is there - in a corporation you support multiple pyramids of leadership. That being said, it takes an intangible set of soft skills to move around with flexibility across projects.

Since I started my PhD journey late in my career, I actually found my industrial experience to be a big plus for Lesson 5. My mentality was to treat my advisor as a senior leader, similar to someone in the ranks of my company's leadership team. Having been groomed to climatize to this type of relationship over decades really allowed me to accept my position and rank relative to my advisor. Now, this doesn't mean that I was subservient, by any standards. Rather, this imbibed me with a professional understanding that helped me to understand the constraints of the PhD student-advisor experience. Mainly, I felt that it was my job to publish my research and to effectively communicate it to my advisor.

Now, if you're starting your PhD journey and things are going rocky with your PhD advisor, I recommend a few things. If the scenario permits, speak frankly to your PhD advisor and express your concerns dispassionately. Of course, if things cannot be worked out and become antagonistic despite your best efforts - escalate to the chair or other parties. Finding a new advisor isn't the end of the world and happens frequently.

Let me state the obvious. Your advisor has a lot of power. In addition to co-authoring every paper you publish (which comes along with reviewing it, providing feedback, and more) he/she is in charge of: (1) chairing your proposal defense, (2) signing off on your committee, and (3) chairing your formal defense. While some of these are milestones in your doctoral journey, it's all the in between - the days and weeks of grinding out experiments, results, and evaluation - that cultivate your relationship.

Before I applied to my PhD, I met with my advisor. I admit, I was very lucky. Because in hindsight, had I not met him and spent the time to share my research interests, I would not have applied. As a matter of fact, I didn't apply anywhere else. I solely sought to pursue my PhD at UMBC because of my advisor.

So, before you blindly apply to PhD programs, do some research to understand who the professor is, what his/her students are like, and if possible find time to speak with the students directly. You may want to ask questions like:

  • What kind of research is the advisor known for?
  • How many students are in his/her lab?
  • What is the advisor's rank in faculty in the department?
  • What is the advisor's management style?
  • About how much time does the advisor spend per student?
  • How would the advisor describe a good working relationship?
  • What is the advisor's expectations on balancing work and research?
  • Has anyone left his/her lab, and why?
  • Where are the advisor's students now?

Lesson 4 - There is more to a PhD Program than the University Name

When I shared my intent on starting a PhD, back in 2020, I was flattered that some academics I knew at Harvard invited me to apply. However, it wasn't for me. I have children and a husband with a demanding job. As a result, it wasn't feasible to uproot my family to a new state, much less without earning an income. The University of Maryland Baltimore County is a great school. It's not Harvard, if you're into traditional metrics that measure "prestige". It's not Ivy and it's not a top engineering school. But my undergrad came from Georgia Tech, which is a Top 5 engineering school, I have a Masters in Public Health from Emory University which is also Top 5 for Public Health, and a MBA for good measure.

My point is - go to the school that is designed for your lifestyle, expectations, and per above, the advisor and research. At UMBC, I was so impressed to see the accolades of many professors having studied at MIT, Princeton, Stanford, Harvard, Fancy School University, etc. with serious work experience in national security, defense, and healthcare. Because UMBC is so close to Washington, D.C. and sits adjacent to many governmental agencies, the breadth of industry experience and academia is huge. You will be at this university for 4+ years. If you select a university just because of its name and not additional factors such as the nature of research, the diversity of faculty, and the potential working relationship of your advisor - you may end up not getting the attention and mentorship you deserve.

Lesson 3 - Learn How to Write Well (and Frequently)

In traditional  and "applied" computer science (like my department - Information Systems) - if you study AI, you will have to publish throughout your PhD career. Some of this I find is due to the perishability of the technology. I mean, look at LLMs even back in 2020 and LLMs now! As a result, my general mindset was that in order to be successful, I needed to have published as much as possible. Now if you end up with an advisor like mine, he/she will force you to push yourself and aim for top AI conferences. For me, this was NeurIPS, CVPR, ICML, ECCV, IEEE conferences, etc.

I failed a lot, but I also scored a bit, too.

Over four years, I luckily (and yes, there is a lot of luck involved), published eleven papers. About 3 of these are from workshops and one is from a doctoral consortium. Only 1 is co-authored. Luckily, I was the primary author of all ten. I think a key factor that I can attest to - and one that I've observed among successful PhD students - is the ability to write frequently and well.

Before I started my PhD, I never really thought that the art of prose had much to do with these mathematically-dense computer science papers. But, I was wrong. There is indeed an art in technical story telling, and definitely, an art in marketing your results. This takes time to understand, as reviewers are expecting something "novel" (ugh, the term) and interesting as they wade through literally dozens of papers. Finding a way to stick out from the crowd and make your results compelling while also maintaining accuracy and scientific integrity is tricky.

When you begin your PhD voyage, you will naturally start to read a lot of papers. This is the first hazing process of any PhD. The first technical papers take hours to read. Then fast-forward three years later and you're a reviewer, and it takes less than 30 minutes. Then fast-foward another few years, and you can get the gist in about ten minutes. This increase in cognitive tempo is likely the result of processing not only the technical content, but also the style of the material. You're not tripping over so many words, complex statements, or diagrams, anymore.

As a result, it's quite obvious that reading is extremely important because it will teach you the style of your domain. You might want to ask yourself these questions as you read:

  • How do authors in my field structure their paper? Is it as simple as introduction, related works, experiments, and results?
  • What's the "wow" factor that draws me to some papers, and turns me off from others? Is it a snazzy name ("Attention is all you need"), the way the authors lay out their results? Specific word choices?
  • What parts of the paper are fast to write (e.g. Related Works), versus which parts are slower (e.g. Network Architecture)?
  • What do the authors really want you to know after reading the paper? If they had a BLUF - what would it be? Did you get that after reading their paper?

Sometimes, the art of good technical storytelling relates to the figures and delivery mechanism of your results. For me, within the field of computer vision, we're known to add lots of pretty pictures. But you can't just add a figure. There needs to be a reason why. The image has to be really large and easy to see. Sometimes you need to annotate it. I didn't learn these things until I read a lot of papers similar to my research and studied what they did.

All in all, to write well requires to write frequently. Whether this means a weekly report to your advisor with incomplete results or an introduction section that you're ginning up in Overleaf - keep writing.

Lesson 2 - Take it Easy on Yourself

I started my PhD in January 2020 and guess what happened that year? Right. To make matters more complex, I moved overseas for a year and a half. So, there were already many forces working against me. But obviously, I kept going. During my four years, I also worked part-time - because well, I need to pay the mortgage. On top of this were my priorities as a mom, undertaking this PhD during very critical developmental years of my kids' lives. To say these four years were stressful, is the understatement of my life. I can't tell you how many times I fantasized about quitting. I've burned out so many times, I must be on Phoenix Bird Number 10, at this point. None of this is to kvetch, but rather to get to an all important point as you sacrifice yourself to the PhD gods - take it easy on yourself.

I recall a very dark time when I was not so mentally healthy. At that point, I actually escaped by starting another project. The end result of that project turned into my first completed scifi novel that I am currently editing. I also remember an 18-month drought when I got rejected six times for a single paper. Yes, rejection stings, but when you get rejected that much - you clearly begin to question why you're even doing this. And worse, you question if you're "smart enough".

Even in the face of incredible stress, undertaking a PhD is also a very weird thing. There's an ancient, archaic element to it. A group of elders debating the merits of your intellectual contributions over a very long period of time where you operate as their apprentice, toiling away on an extremely narrow research problem. Know this. When you graduate, you will not be the world's computer science expert. You may not even be enough of an expert in your domain (e.g. computer vision). You will, however, be an expert in that thin slice of research.

This isn't to degrade your experience by any means. My point is that all of this PhD stuff isn't worth you annihilating your mental health or destroying your life (and others around you, in the process). When you reach that point of burn out, which absolutely will happen, consider the following:

  • Where can you escape? Such as a hobby like sports or music. For me, it was writing my novel and running. Can you turn this into a routine? For example, I run almost every morning for several miles.
  • Can you speak with your advisor and schedule a break in your calendar? A few weeks, for example. And negotiate on completely "off" days where you do not respond to email or phone. (As a side note, I would highly recommend learning via Lesson #5, whether or not your potential advisor is a micromanager. If so, steer clear.)
  • Do you need to take a few months off? Yes. I know PhD students who have multiple personal and professional responsibilities, and took an entire year off. If you think this is an option for you, make sure you check your university's graduate school policies.
  • Above all, talk to someone about your mental health. If it isn't a friend or family member, please reach out to someone professional. Again, it's not worth it. It's literally just a PhD - keep that in perspective.

Lesson 1 - You are the Master of your Destiny

No one will hold your hand during the PhD.

Perhaps this isn't the case with all PhD programs, but for me, entrepreneurship was highly rewarded. The only way I got through my PhD was by treating it like a job. I had to mentally design my own scope of work in order to assign myself a task for the week. Sometimes I wouldn't even know what to do. I just had to trust that Catherine A could lead Catherine B as her boss.

Everyone is different. But for me, I didn't analysis-paralysis myself to death. I live on a "just do it" mentality since my time is extremely precious. I didn't have time to make lists in perfect kanban boards or anything like that. I basically developed a project management philosophy that worked for me. But it hinged on these basic, tactical tasks:

  • Where is my current research leading? In other words, is there a conference I should submit this paper to? Great. When is the deadline? Cool. That's my deadline.
  • Now, estimate how long the experiments will take. Write the code for it. Review the results. This should take x amount of weeks. Meanwhile, write weekly reports for what the results mean.
  • Start the Overleaf draft. Write a pretty awful intro and related works section. Put in the headers for everything else. Go back to item #2.
  • Results are ready. Write them, make pretty tables and figures. If for whatever reason I couldn't get this far in time, then be honest about it and let my advisor know. There are other conferences.
  • Schedule meeting with my advisor and present the results. Get his feedback. Submit. Rinse and Repeat.

Because I also worked part-time, I had to balance the PhD work with the work that paid my bills. As a result, I had some very intense weeks of coding and writing. After the paper was submitted, I allowed myself up to two weeks to rest and recoup before I did it all over again.  Sometimes this meant binge-watching Euphoria or GoT (yes, I didn't get into that until 2021).

I revolved my part-time corporate work around this lifestyle. If you're planning on being a T.A or do lab work paid by a NSF grant, you'll likely follow a similar routine. Just be ready to live in chaos for a few years.

Lastly, if you decide to quit after deliberating all the reasons not to - then do it and don't beat yourself over it. When I started my PhD I knew that I was non-traditional in having started it pretty late in life. But, man, I'd regret it for the rest of my life, if I hadn't at least tried. Hence, you really do only live once. You can alway return to a PhD program. I don't think PhD programs are going anywhere. But I might recommend that if you do decide to quit, that you try to get a Masters degree out of it, or something else that offers value during the time y0u were there.

Final Thought

I can't leave this post on a downer like that.

I guess if you really made it this far, you're either interested in a PhD program or are being drawn back to your captors, reminiscing in your masochistic journey as a former student. (J/K).

I'll end by saying that there are many advantages of having (nearly) completed the PhD program. Here are my greatest hits:

  • You learn how to propose a very difficult research problem, and actually make strides to solve it.
  • You will be astonished at the range of how far you can be intellectually pushed.
  • No high can match the rush of publishing your first paper.
  • You are contributing to science through incremental, baby steps.
  • You cultivate a healthy dose of skepticism.
  • Whatever work ethic you began with, will go intergalactic by the time you're done.
  • You just might have a hand in solving an incredibly important problem that will eventually change society for the better.