The Professorial Pivot

(Or how I’ve learned to stop programming (mostly) and embrace being a sometimes designer, sometimes manager, sometimes pep talk deliverer.)

In the fall of 2009, as I was preparing to graduate, I went on a series of self-funded university visits 1, each consisting of a talk and multiple one-on-one meetings with faculty members. While I had previously met many faculty at conferences, this was the first time that I had encountered them in their more natural setting—​in their labs and offices, teaching and working with their graduate students. And so it was the first time that I was exposed to what I now hear fairly regularly, with varying degrees of regret and resignation, particularly from colleagues working in computer systems and networking: "I wish I had more time to write code. I’m just a manager now."

I really liked to program, to build things, to hack. (And still do.) I was also interested in becoming a professor. But I didn’t interpret those complaints from systems faculty as a warning about my own prospective career path. Instead, I remember feeling annoyed. "You knew what you were getting in to", I thought. "Do your job 2." I had an advisor who liked to try to help out with our projects, and it didn’t always work out well. As a student, finding bugs in your advisor’s code was naturally a bit nerve wracking, and at times his internal programmer would cause him to exercise an unhelpful amount of low-level control over the code that his students were ultimately responsible for writing and maintaining.

I’m remembering these feelings now, having recently gone through the transition from programmer to manager myself—​my pivot, as the industry jargon goes. Like others before me, I’m finding leaving my development role behind both frightening and disorienting. But I think it’s a good thing for my group, and so I’m trying to adapt to my new role.

Hit The Ground Hacking

Transforming from Professor Hacker to Professor Manager took longer than I would probably have expected 3, because for the first few years after arriving at UB I found plenty of things to hack on. I wrote code to run our PhoneLab smartphone testbed and a suite of tools to analyze the data we collected. I wrote code to automate submission and grading for Harvard’s OS/161 assignments, allowing me to assign them in a class with an order of magnitude less instructional support. I pitched in wherever needed on a bunch of our early projects, helping with things like building simulators and writing graphing scripts.

Looking back, my initial years of hacking can seem like a reasonable response to events. We quickly got a lot of money to build a smartphone testbed, and so that needed to happen—​even if I hadn’t yet assembled students that I trusted and had the required development abilities. PhoneLab is also really something akin to a production system, not a research prototype, and it had to work at the risk of irritating and alienating our testbed participants. With my course, I threw my hat over the wall and there was no way to make it work without devoting a lot of early effort to automation. And the course was another thing that really had to work, since it was being used by tuition-paying students. Finally, working with graduate students that are just getting started, you find it natural and times to lend a hand when they get stuck.

Being set down to a challenging task in a new environment, it’s easy to reach for what you know.But a lot of it was driven by fear. Being set down to a challenging task in a new environment, it’s easy to reach for what you know. I knew how to program. I didn’t know how to manage an engineering team. And so at least for a while hacking allowed me to feel like I was making some progress. I suspect that this may be a common experience, particularly for systems faculty, and particularly for faculty that don’t get a lot of experience managing other students while completing their PhD—​which I bet is most of us. But ultimately, the urge to fall back on familiar strengths can hinder the development of the new skills required by new situations. I was lucky that a change was around the corner.

Suddenly, A Pivot Comes

This article claims that "nearly all engineering managers remember their pivot—​that moment when coding was no longer their number one priority." I definitely do, and it when the transition happened last semester it felt like overnight I suddenly had no development role on any of my group’s projects.

Initially this felt exhilarating, because the root cause was the fact that my students were working so hard on so many cool projects that it was impossible for me to keep up. It was like I had spent years pulling from the front desperately trying to get things started, and then all of a sudden I’m being run over as everything takes off because of a bunch of people are pushing from the rear. Even feeling a bit shaken and more than a bit left behind, I still had a distinct urge to cheer 4.

Having gotten out of the way, I’m finding that things just happen.Having gotten out of the way, I’m finding that things just happen. We did a bunch of projects this semester where I contributed zero lines of code and yet new systems got built, deployed, and evaluated. Having never previously managed a group of developers that I have grown to respect and trust, it’s pretty awesome. And at the same time I know that my students are enjoying themselves, because they like hacking too and they get to the do those fun bits that nobody lets me meddle with anymore. So they’re having fun, I’m watching interesting things get built, and everyone’s happy.

But Where’s My Paddle?

But my new reality is also a bit bewildering and puts me into a position where I’m challenged to find new ways to contribute to my group’s future success. I am well aware that, while I have somehow assembled a fantastic engineering team, I am not yet a good engineering manager. That’s something that I’m still learning how to do. At the moment my approach to managing their development process is probably a bit too hands off, and we should probably be doing things like code reviews in addition to having high-level design discussions. At least so far the hands-off approach seems to be working, but I think I can do it better—​and will need to, soon.

On the other hand, my early days as Professor Hacker did help me set up my lab in a way that allows things to get done. Mainly, I don’t force my students to endure regular fixed-length meetings 5, the main purpose of which is to make me feel like I’m contributing to their project. While the initial purpose of my ad-hoc meeting stance was to let me focus on programming, it works equally well for letting my students focus on programming. But I will admit that without writing code, I do sometimes feel a bit like I’m not contributing enough to what we do. Maybe I should hold more meetings!

I also have had to learn to be more patient. When you’re less involved with the development process, it’s easy to get frustrated at why it is taking longer than you anticipated for something seemingly simple to get done. Sometimes this is because the student made some kind of boneheaded design decision or implementation choice that just needs to work itself out naturally. But more often than not it’s because what seems simple at a high level ends up being complicated to implement well. So although my patience as a manager has been tested recently, here another aspect of our working arrangement helps. Because I’m in the lab all day with my students, I know that they’re working hard. And as long as they keep working, they’ll either discover and correct their bad design decisions, or finish implementing the hard thing that seemed simple to me. That makes the waiting easier. It also helps that I have more time to manage the process and can make sure that we start early enough to meet important deadlines.

Having such strong programmers to work with also challenges me to develop new ideas for us to work on, and one of the things I worry about frequently now is coming up with novel systems for Anudipa (or Guru, or Jinghao, or Nick, etc.) to build. One of the ways I’m responding to this challenge is starting to think of myself our group’s chief designer, although I expect all of my students to also play creative roles in our projects. (Once a student is able to independently design, implement, and document their own new systems, it’s time for them to graduate.) I usually tell students that systems research is 10% coming up with ideas and 90% getting them to work, and so spending less time on the 90% should leave me with more time for the 10%. Given that we don’t have the automatic legitimacy of a group at MIT, this is probably an important thing for me to focus on.

Unexpected Benefits

In addition to more time to focus on these other aspects of being a good teammate for my students, I’ve found the shift away from programming to have two interesting and somewhat unexpected benefits.

First, I find myself enjoying reading research papers much more than I ever have previously. Part of this is probably just a bit of temporally-overlapping maturing on my part, but I think part of it is also not having the little voice in the back of your mind when you’re trying to read a research paper whispering "But when do we get back to building our system?" That’s made it a lot easier to appreciate other people’s work, which is a critical part of both developing new ideas and staying on top of what’s novel in my new role as designer.

Second, I think that I’m doing a better job of writing portions of our papers. (We’ll find out for sure once reviews for some of the half-dozen or so papers we have under submission start coming in.) Of course, not having to simultaneously build and document the system leaves more time to write, but at least for me the improvement doesn’t seem only due to the extra time. My hypothesis is that programming a system puts you in a certain kind of mental state that makes it hard to write eloquently about the same system ten minutes later—​or maybe just hard to write with the level of sweeping grandiosity that seems necessary to produce a winning introduction. You’re just thinking too much like a hacker, which makes it easy to write (or "hack" together) prose that’s uninspiring and often confusing.

I actually struggled in a way that I had never experienced before writing up parts of one of our group’s submissions this year, to the point where I found myself wandering around campus aimlessly feeling kind of like I was having a nervous breakdown. (This is not usually what I do at work.) I think it was because I was actually trying to get the text and the design arguments right, whereas previously the programmer-writer version of me just got them "to work". In any case, since writing is, for now, still something that my students let me do (sometimes), writing better is a welcome development, even if it requires long walks outdoors.

Getting the Balance Right

Despite feeling a bit wistful about my bygone days as a developer, I wouldn’t trade my team for anything and so I won’t complain excessively about my new role as our group’s manager and designer. While I’ll continue to have a direct role in things like keeping the lights on and the paychecks coming, I anticipate that more and more my job will be to help keep my talented students creative, productive, and happy.

But in other cases I think it’s more important for early-stage faculty to establish an example of the kind of graduate student they want their students to beLooking back, I do wish that I had trusted my students more a bit earlier along and let them take the lead a bit sooner, particularly given how well that is working out now. But I’m also glad I didn’t make the mistake of trying to be too managerial too soon. That may work fine when there are already strong students waiting to be led—​at top-tier schools, or when you’re joining a department in an area of established strength populated by strong colleagues. But in other cases I think it’s more important for early-stage faculty to establish an example of the kind of graduate student they want their students to be and then wait for their team to materialize. After all, everyone knows what they say about a leader without any followers.

I am plotting to find a way to start writing some code again. Part of the reason is to follow Jinghao’s advice, who reminded me to "stay sharp." Part of the reason is to continue to participate in the hacker culture that attracted these strong students to my group in the first place. But part of the reason is simpler: I still like to hack.

So what to work on? At this point I’ve decided fairly categorically not to take development roles on our research projects, since this is something that on every level is better left to my students. Happily, there are still a lot of non-research development tasks left for Professor Hacker to work on. Like building this website, and updating our PhoneLab infrastructure and tools. I have some new ideas I want to try out in the classroom that will require some new online infrastructure, and I’ve been promising myself that after over a decade I would finally repeat the OS projects I assign to students 6. So there’s enough to keep me busy and give me the opportunity to maintain my skill set—​keep my code hand dirty, as we say.

And while it may not count as coding, I can always write on our new blog.

Built by the metalsmith-blue Metalsmith pipeline.
Created 5/13/2015
Updated 8/17/2017
Commit ed24937 // History // View
Built 8/17/2017 @ 19:29 EDT