Simulating a PhD

The Secret to Operating at the PhD Level

Seems like job hunting in QC is in the zeitgeist recently

¹

, because both Olivia Lanes and Michał Stęchły have written recently about various facets of getting a job in quantum computing. If you’re an outsider like I was, you should give these both a read.

Serendipitously, I’ve been thinking about a related topic, namely what to do if you’re a recent hire with a BS who is likely to be competing with fresh PhD hires in 3-5 years

²

.

Let’s set the stage: you’ve just started your dream quantum computing job, fresh out of undergrad. You have a BS in something relevant or relevantish, perhaps Physics or Electrical Engineering. Maybe you even did a few internships in a quantum computing focused academic lab?

But now it’s a whole new ball game. Many of your new colleagues have PhDs in quantum computing and have many additional years of experience. Every day, newly minted PhDs join the company and you can see that they’re showing up pre-loaded with a ton of knowledge. There is a clear knowledge gap between you and them, but that’s OK, for now.

The question is can you, in 3-5 years, build yourself a reputation such that you input on technical matters is taken as seriously as your PhD having colleagues? Or, in other words, will you be able to do their job in 3-5 years? Simply showing up everyday and doing your job will not be sufficient to achieve this. I’m sorry. That’s just how it is

³

.

Luckily, I am here to tell you the one weird trick to solving this problem!

What Is Your Job?

Your technical leads should have a pretty clear idea of your role and responsibilities, which they should, of course, explain to you. But how should you think of your actual job? One way of conceptualizing what you do at work is training your boss to trust you.

Every day, you are working toward training your boss, or team lead, or VP, or whomever, to implicitly trust your expertise, your skill, and most importantly, your good judgement. The fastest advancement I’ve observed, not only in title and compensation, but also in respect and reputation has been among people who are the answer to the question “Who can solve <Insert Problem Here> for me?” for the largest possible class of set of problems.

Every time you display good judgement, be it by vetting your data, anticipating objections, questions, 2nd order effects and so on, you reinforce the idea that you are a highly competent professional who can be trusted with more responsibility

⁴

.

Again, showing up every day, doing your job and then going home is unlikely to get you where you want to go. Remember, in 3-5 years you are competing with people straight out of PhD programs, who have spent that time developing the expertise in your chosen field. The PhD isn’t magic, but it is a strong filter for the kind of person that is comfortable exercising reasonably good judgement in environments of high ambiguity and uncertainty

⁵

.

This quality of judgement comes from deep understanding of both what is supposed to be happening in the experiment/simulation/hardware and what is actually happening. It distilled from challenging grad school experiences, where progress is won by slow, grinding perseverance. So, to get to the same point, you need to emulate some, but not all, elements of grad school.

Closing the Gap

What does it mean that your colleagues have PhDs? As I alluded to above, PhDs have built deep expertise by reading vast amounts of literature, they’ve demonstrated the ability to act effectively and independently, to control a project over the course of many years to completion. They’ve put in thousands of hours of work to develop their skills and understanding.

Let me give you an example of what I expect newly hired PhDs to do. I have, on multiple occasions, asked new, PhD-having members of my team to improve our numerical models to capture some additional physics we were missing before. Here is what I gave them:

• A brief outline of the state of the model.

• A description of what additional things the model should do.

• A bunch of papers about the relevant physics.

That’s it. Then they went and did the thing. They read the papers, they consulted with other members of the team, they came to me to ask clarifying questions, they sought out experts for more information, they reproduced results from the literature to vet their understanding, and then applied their knowledge to our specific system. Most importantly, none of them did their PhDs on the topics I asked them to learn. Because of their training, because of the years they spent grinding through inscrutable papers and chasing down bugs and weird numerical artifacts, they were able to excel at their assigned tasks. There are probably a small amount of BS holders who could do the same, sure, but the difference is I expect any

⁶

PhD who walks through the doors of our building to be able to execute on this.

This is the gap you need to close, and it is doable. Unfortunately, by trading the opportunity cost of going to graduate school for a Real Paycheck, you also lose the freedom to struggle, fail, and work slowly to do something simple, because now the corporation isn’t making money when you’re not making progress. It is unlikely you can spend a week on some derivation you don’t understand while your numerical models gather dust or your fridge sits idle.

How to do it?

If I had to reduce a PhD to a few key features, it would be these:

• Can independently and thoroughly vet collected data. You shouldn’t be showing your colleagues or bosses bullshit data, and you should know the difference between bullshit and legit data. You should know which questions to ask to determine the difference.

• Has deep knowledge of the relevant literature and can efficiently read and evaluate new papers. Knowing the literature helps with item 1, but also provides context for all of your efforts and becomes a source of new ideas. You should be able to break down sections you don’t understand, find the core problem and work it out.

• Can act independently to resolve difficult/ambiguous/generally weird technical problems. In ‘‘‘deep tech’’’ even mundane problems are weird, present odd symptoms, and generally don’t want to be fixed. You need to be able to diagnose and resolve these on your own, or do enough that it’s not embarrassing when you call in some help.

  ⁷

The good news is that, ostensibly, you can execute on the first and last on-the-job, as they are closely related. If you understand your system well enough to fix it under most circumstances, you should understand it well enough to know how the data is produced and what makes the data trustworthy or not. The downside is that the easiest way to understand how to fix something is to break it, something for which there is great tolerance for in grad school, but maybe not as much in industry.

The harder part is developing familiarity with the literature. Reading papers is a skill, one that takes a while to develop. There’s still hope, though! It’s possible that your new workplace has a journal club. If it does, you should go and participate. Read the papers, even if it means you need to the reading on your own time at home. Ask lots of questions, make tons of notes.

Here’s a set of questions

⁸

that can act as a reasonable barometer for your expertise:

• Before collecting data, what do you expect it to look like?

• Once you collect the data, it is likely wrong. How can you confirm that it is wrong?

• If you fail to confirm wrongness, then maybe it's right. What does the data imply? Can you explain all features of the data?

• If someone were an asshole at a conference, what questions would they ask about the experiment to crush your spirit?

Having good answers to these questions is a sign that you know what you’re doing. Further, for the kinds of experiments/simulations/other tasks you’re likely to be doing as an employee of Big Quantum Inc, you will almost certainly need to intimately understand your instrumentation, your device, and much of the literature relevant to your task.

Graduate School Or Professional Development?

Getting to a PhD equivalent level of proficiency in QC without actually attending a graduate program is going to be pretty hard and take some real commitment. Your employer might have access to useful professional development resources that could help (ongoing education benefits, etc), but such benefits and attitudes toward continuing development definitely vary between companies. It’s worth asking about this kind of thing during your interview!

The other option, of course, is to eat the opportunity cost and attend a graduate physics program to get a PhD in a quantum computing or quantum information lab. One thing I suspect is true is that a few years of industry experience will provide both useful experience that might make you a more attractive PhD student than someone fresh out of undergrad, and monetary savings to help defray the impact of the smol graduate stipend

⁹

.

Note that PhD program represents a substantial opportunity cost, and the length of the program represents a risk that you might graduate in time for Quantum Winter to hit, causing the industry to contract substantially. I would only recommend this route if you find that you really love the subject.

Some particularly motivated colleagues of mine ended up taking advantage of our corporate continuing education benefits to attend a Master’s physics program at night. This seemed really stressful, but they are extremely good at their jobs and their technical knowledge is top-notch. Of course, it might just be that the kind of person willing to do a nighttime Master’s is also the kind of person who would have excelled without the degree.

Indeed, a major benefit of hiring PhDs is just that they’re self-selected for independence and perseverance, which are important ingredients for success anywhere. It might seem like I’m saying that the people who were going to succeed anyway, are going to succeed, but perhaps, if you are not such a person, this post can help nudge you toward power and glory, rather than complacency and stagnation.

---

1. This was true when I first wrote these words. It has been many weeks since then.

2. I choose that interval because PhDs take something like 4-6 years, but it’s common to spend a year or so doing coursework before getting actual lab experience.

3. This is mostly because an industry job is not a PhD program, and cannot usually be treated as such. They’re fundamentally different activities with wildly different ultimate goals.

4. And therefore deserves more: recognition, respect, compensation, etc

5. This is generally true in technical/scientific matters. It, sadly, doesn’t necessarily generalize to our personal lives.

6. Any STEM PhD, of course.

7. It’s still going to be embarrassing, sometimes. Some people just have the “Working Experiment Aura” and their mere presence will make your problems go away. You’re not crazy, it’s a thing.

8. They are focused on experiments, but should be straightforwardly generalizable.

9. Unfortunately, I have no contacts who have even attempted such a feat, but if you, please let me know the result by commenting or shooting an email to quantumobserverblog (at) gmail (dot) com.