By Victor Zhong, August 20, 2025
A year ago, I started my journey as an assistant professor at the University of Waterloo. I could try to write a guide on how to be a professor, but the truth is, everyone's path is different. Instead, I want to share a personal reflection on my first year: what I learned, what surprised me, and why I am doing this in the first place.
The notion of building a lab was completely foreign. I had two things going for me. First, Waterloo graciously allowed me to take a gap year. Second, Microsoft graciously allowed me to prepare for professorship during my postdoc. This gap year was invaluable, as I was able to recruit and admit students and write my initial grant. Having that preparation time before officially starting my job made things significantly easier once I arrived in 2024.
How do you hire students? I asked this question to several faculty friends before I had to hire any. The answer I got the most was "trust your gut feelings." On the one hand, this is a terribly dissatisfying answer because it really isn't actionable. On the other hand, much like faculty hiring, graduate student hiring is a random process with very high-variance signals. I use the usual factors to evaluate students: their research experience, prior publications, how deeply they understood their own work, and recommendation letters. After prioritizing students who meet these criteria, it's hard to distinguish between them.
The final step for my lab is an interview, which simply aims to evaluate cultural fit. I ask students why they want to work with our lab, what kind of research they're interested in, and what they want to do after graduation. I am also transparent about the pros and cons of working here. We're a new lab, building for the long term. This means students won't have a lot of mentorship from senior students or a pre-built publication pipeline to slot into. Instead, they will identify and lead their own research projects: a very daunting task for new graduate students. The interview is a forced "gut feeling" check on how much I think I would enjoy working with the student, how much my other students would enjoy working alongside them, and how much they would enjoy being part of our lab.
I have had the great fortune to be a part of multiple great labs at Stanford (with Chris Manning) and at the University of Washington (with Luke Zettlemoyer). From these experiences, I've learned two important things:
It takes time to build a lab. You have to be very patient, and more importantly, very consistent. Both Chris and Luke exemplify these two characteristics.
Lab culture is very important and very brittle. You have to always try your best to find the right fit. It's very hard to build culture, but very easy to break it. I am still learning how to evaluate "fit". It definitely seems more art than science!
How well did this work? I am generally very happy with the way things worked out. The R2L lab is a very cohesive group of students who want the best for the lab. They are active in discussions, brainstorming, and volunteering to help each other out. We're almost exactly a year in, and it's been a very engaging learning experience for all of us - myself included!
Writing the first grant was scary. I remember sitting in the NYC Microsoft office one afternoon during my postdoc. On the left of my screen was a blank Word document. On the right was the incredibly long specification for the NSERC Discovery Grant. I thought to myself: what the hell have I gotten myself into?
Since the Discovery Grant, I've written about 10 grants. For Canadian PIs, this is probably a lot, as most of my colleagues write about 2-4 a year. For me, however, I wanted to practice. After a year, I'm happy to report a significant increase in the quality of my proposals, as well as the corresponding yield. The biggest change? I stopped writing for myself and started writing for the reader. I am much more intentional about describing and addressing what the reader wants and needs, as opposed to mindlessly explaining the details of my work.
What happens once you get a grant? This depends on how you want to build your lab. My intention is to build for the long term. Consequently, a bulk of my budget is for long-term compute resources like servers and GPUs. The remainder funds my students. Deciding how to spend money is scary as it has long-term implications for your group and your research. However, budgeting is an important exercise. We have to put in a lot of time and effort to do impactful research, and we also have to acquire the resources to enable our students to do that research. Without results, it's difficult to acquire more resources; without resources, it's challenging to do (most empirical) impactful research. Balancing resource acquisition and research activities, both of which demand significant time and mental energy, has been a key challenge. That being said: if you work in AI, you'll probably have more money than you think will. Fund-raising is a skill that will improve as you keep working on it. Your hardest grants are likely to be the first ones you write.
The most common answer when I asked faculty friends "what should I watch out for" was about time management. This topic, similar to building a lab, is very personal. Therefore, time management advice is typically dissatisfying because it's overly general. After a year in, I've learned to adopt a rather simple guideline: if I enjoy something more, I will try to allocate more time to it.
Of course, many constraints affect the implementation of this guideline. Professors have hard deliverables they must meet, such as teaching, grading, and research meetings. On top of this, there are external and internal engagements, such as departmental service, institutional service (Vector Institute), national service (CIFAR), and international service (e.g., conference chairs). Finally, there are activities in which you can devote an infinite amount of time, like research projects and grant writing, because there's no finish line.
I prioritize according to my guideline given these constraints. I enjoy brainstorming open research ideas, teaching, community-building, and learning/applying state-of-the-art technology. I don't enjoy meeting logistics, emails, and grading. My strategy, therefore, is to make the first type of activities basically immovable on my calendar, while increasingly automating the second.
For example, my students have priority over my calendar—I have one weekly meeting with each student, in addition to our group weekly meetings. If I'm teaching, I block off time for classes and office hours. I also allocate a few days per month to engage with companies to present what we're working on and learn about the open problems they face in practice.
For the other necessary tasks, I try to automate them as much as possible:
Scheduling: A amazingly large part of the job has to do with scheduling. I use a bookable calendar from cal.com to avoid scheduling headaches and endless back-and-forth emails.
Teaching: I am in the process of making auto-gradable assignments for my Intro to AI class. This takes a lot of effort upfront but will hopefully reduce the amount of time I and my TAs have to devote to grading in the long term.
Tenure and Reporting: One very helpful piece of advice from our new faculty onboarding was to allocate some time each week to track what we've done that potentially contributes to our year-end progress reports. I maintain a running document in Overleaf using the university template. Every Monday, I allocate at most 30 minutes to update it.
Weekly Research Meetings: My meetings with students are very structured. This is a minor shock at first, but it becomes very helpful as we both know exactly what's expected. My meetings are structured as follows: first, the student states what we agreed to do last week. Second, they talk about what they've done and ask questions, and I will talk about what I've done and ask questions. Third, they propose what to do next week, and I offer suggestions. This is how I did my PhD with Luke, and I've found it very helpful to explicitly communicate this to students when they start.
After all these necessary items are scheduled, I fill my remaining time with things I enjoy. The largest category is community building, such as speaking at conferences/workshops and Vector/CIFAR events.
A year into the job, one question that comes up much more frequently than I thought it would is "why are we doing this?" The landscape of research has changed significantly over the last decade. The first part of the decade was characterized by optimism and an abundance of resources. The last few years have been characterized by resource prioritization, probably due to economic pressures. I think it's more important than ever to be doing this job for the right reasons. The following is not a moral judgement on right vs. wrong reasons, but my assessment of how good of a reason these are in today's environment.
I think the following reasons are not as good as they used to be:
Prestige: The professor is, frankly, not as prestigious of an occupation as it used to be. Under today's economic pressures, jobs that bring direct economic value are often seen as more prestigious, such as the research scientists who work on Gemini or GPT. One can certainly make a case for professor making economic impact, however most of this impact is very difficult to measure due to its indirect nature (unless, of course, the professor starts a successful company).
Influence: I am curious about the perception that a professor has influence. While top professors have significant influence, does the general professor have any more influence than, say, a lead software engineer at Google? My impression has always been no. As a professor, you have to convince grant agencies and companies to support your research. You have to convince good students to work with you—students who would almost certainly make more money by working in industry. In practice, the professor has very few hard leverage, and soft leverage means little in the face of hard challenges (e.g. political instability, economic woes).
Job security: In recent years, several institutions have stopped giving professors tenure—a trend that I suspect will continue, at least in the United States. We have also seen plenty of layoffs and buyouts during recent budget crunches. A professor is still a high-security job, but it's undeniable that job security has seen a steady decline in recent years.
On a more positive note, I think the following remain great reasons to be a professor:
The opportunity to work on open-ended research: We work on many (what I consider to be) creative ideas at the R2L Lab. I honestly find this very exciting. Moreover, I have a group of students who also find this very exciting. I find it unbelievably rewarding to first, truly believe that what we work on is very important and unique, and second, work with students who share this belief and work hard to make our vision a reality. When I worked at MetaMind, somebody told me: the thing with working at a startup is that you might never want to work for a big company ever again. He probably described my subsequent trajectory pretty well, however neither of us thought that I would be at a university instead of in another startup!
Community building: One of the main things I've learned about myself is that I enjoy community building even more than I thought I would. On a micro level, I love helping students who put in the work succeed. Some great things that happened this year is include Tianbao - Tao's student who I've worked with multiple times and came to know well - getting a job at OpenAI. Second, I don't mean to take any credit for any of Sewon's success, but I am very very happy that she got a job at Berkeley. Third, I also managed to place three of my students in great internships where they can do open research this summer, which I think is probably a difficult thing to do for a first-year lab. On a macro level, I love building our research group. At last year's end, I asked the group what we can do to improve this year. We got some great feedback and we're implementing them right now. I love building our department. I've been honoured to represent the department a number of times this year, and I will look for more ways to help as my group stablizes. I love building the Canadian AI system. I have thoroughly enjoyed being a part of CIFAR this first year, and I appreciate the CIFAR community very much. There has been so many opportunities, facilitated by CIFAR, for us to engage to brainstorm ways to help out both our local communities and the country. In the coming year, I'd love to spend a bit more time learning about the local Kitchener-Waterloo AI community to see if I can help out here. I'd also like to push for a more tight-knit AI group at Waterloo, given the large number of new people.
Ultimately, doing the open-ended research I love to build my communities is why I love this job. It's an opportunity to create something that lasts, and that's a goal worth fighting for.
Looking back on my first year, the lack of major surprises was the most surprising thing of all. This is a testament to the incredible help I received from friends and mentors who shared their honest experiences with me. I'm profoundly grateful to Danqi Chen, Tao Yu, Yoav Artzi, Karthik Narasimhan, Tim Rocktäschel, Daniel Fried, and Jesse Thomason for their candid reflections. My thanks also go to Caiming Xiong, Richard Socher, and Ed Grefenstette for teaching me about managing creative research - much of which has to do with how to effectively work with people. Finally, I cannot emphasize how much I appreciate Luke Zettlemoyer and Chris Manning’s mentorship. Neither Luke nor Chris are the type to tell you what you should be doing or what is the right way to do things. However, they are great thinkers and great builders. I think by growing up in their labs, I developed a sincere appreciation and joy for science and community, reasons that have always been true, and always will be true in the future.