Reflecting on my first two years as faculty at CMU
Part 1: Postdoc year

# Preface: Taking stock *Date: Nov 30th 2024.*
I'm sitting by my fireplace after a nice and relaxing Thanksgiving with some of my family members and loved ones. Everyone has gone back home, and I have a full day to myself. Normally, I'd probably consider working at this point, since I took a couple of days off, but I decided I wanted to write this post, and reflect on my time so far at CMU. ### Introducing this blog post series After my blog post outlining how my job search went, I realized that sharing my experiences could be helpful to people. But I also realized it was helpful to me to write my story, and reflect on what I have gone through, my feelings, my misconceptions, my lessons learned, my internal narratives, etc. It helped/helps me to reprocess some events and realize maybe I am okay, maybe I'm not a failure? ### Career-wise I'm in my third year at CMU in the LTI department. I just went through the first "checkpoint", i.e., 3rd year (internal) reappointment, and I have just been recommended (last month) for reappointment, which is great 🥰. ### Advising- and student-wise I like to say that I currently have 8 PhD students. Technically, the picture is a little more complicated, since one of those is a master's student (MLT) that I am very much hoping will stay on as a PhD. Plus, one of my other students is a PhD student at MIT, I only co-advise. Four (half) of my students are co-advised, which does dilute the financial responsibility I have to fund everyone (see more later). In addition to PhD students, I have advised / am advising quite literally dozens of professional master's students. CMU has master's programs that have research components (directed research, capstone, etc.), so there are often teams of master's students. Additionally, some of these students are working closely with some of my PhD students, which means I don't need to advise as closely. However, I do have a couple of master's students that I directly advise. ### Grant-wise This is weird one to write about, but despite my constant feelings of anxiety, I believe I may be financially okay as of now (for now?). According to my internal grant writing tracking Google sheet, I've written or been a part of about 35 grant proposals so far. Only two of those have been NSF grants (one accepted but I wasn't PI, one rejected where I was PI, and one still pending where I am PI). One is a DARPA AI Exploration that I am PI on. Most of them are actually industry gifts or sponsored agreements, which are smaller (100K or less). For context on my feelings towards funding, PhD student tuition and stipend at CMU are *expensive*: to cover a student cost I have to raise between $100-120k a year depending on grant source, overhead allowed, etc. The tricky thing is, unlike in many other schools (e.g., UW where I did my PhD), TAing does not cover a student's entire semester; I'd have to double check, but it covers less than half of a student's tuition and stipend. Plus, students only get roughly half of a year's worth of coverage from the CMU School of Computer Science, unlike some places which offer a full first year covered by the graduate school. ### Teaching-wise Setting aside how I got here, this year I'm teaching classes that I've either solo or co-taught before. I'm teaching my "Ethics of NLP" class (11-830) in the Spring (solo), and just wrapping up teaching an 80-person data science seminar class (11-631) this Fall (also solo). I feel like I've gotten into the groove, and actually enjoy teaching. # Part 1: Some accounts from my postdoc year [Last we left off](./6-job-search-2020.html), I had just finished my job search and defended my thesis. In July 2021, I started my postdoc gap year at Ai2 with one of my advisors Yejin Choi. I probably should have written about this earlier, cause that feels like a lifetime ago now 😅 oh well. High-level timeline: - *July 2021*: Start postdoc gap year before CMU - *July–November 2021*: Wrap up some of my own research, start collaborations with Ai2 folks. - *December–February 2021*: Participate in PhD admissions at CMU, conduct phone interviews, go to in-person visit days in Pittsburgh, etc. - *January-February 2021*: Experienced a lot of personal trauma & grief (breakup with boyfriend of 1.5 years, forever stop talking to my mother), so took it slow at work, focusing on healing and fostering my close friendships in Seattle. - *February–March 2021*: Started working on my first small (3-4 pages) grant proposal for Meta. - *March 2021*: Started a reading group around "[*An Everyone Culture: Becoming a Deliberately Developmental Organization*](https://bookshop.org/p/books/an-everyone-culture-becoming-a-deliberately-developmental-organization-robert-kegan/7343719?ean=9781625278623)" with junior faculty at CMU. - *March–June 2021*: Worked on more research and collaborations. Did a little more grant writing. Submitted to EMNLP 2022. - *May 2022*: Visited Pittsburgh to look for housing. Decided to get a luxury apartment to ensure my living situation was comfortable and easy. - *June-August 2022*: Took some summer interns, took work a little slower as I spent a lot of time with my friends in Seattle saying goodbye to them. ### Finishing up PhD research and advising junior students My priority at Ai2 was to conduct exciting research, and essentially take a little bit of a breather from the stress of the PhD and job search. I led a couple of projects during that time that I managed to finish to completion before the end of my postdoc ([*Annotators with Attitudes*](https://aclanthology.org/2022.naacl-main.431/) by the October 2021 ARR deadline, [*Neural Theory of Mind*](https://aclanthology.org/2022.emnlp-main.248/) by the June EMNLP 2022 deadline). There were other lingering projects that I was wrapping up, including a journal version of my work on measuring narrative linearity in recalled vs. imagined stories which had been dragging on for two years, as well as various collaborations (e.g., the infamous Delphi project, other collaborations with amazing senior PhD student Hyunwoo Kim). I did start a project in the early 2022 that I unfortunately was not able to finish, which my PhD student Xuhui ended up taking on as his first research project. I also started some advising-oriented research projects with some junior students (undergrads, master's), which led to some good preliminary results that we published in NLP workshops. Since those works tend to be more "in progress" and "proof-of-concept", these workshop paper projects provided a nice foundation to give to my junior PhD students as guidance for what they could work on next. One thing I realized during that time—frankly I wish I had realized this during my PhD—is that it is really hard to multi-task projects. By that I mean, it's really hard to make steady progress on multiple research projects that you're leading. I kept disappointing myself by not being able to check items off my to-do list at the end of the week. I forget how I reached this decision, but I one day decided to just stop trying to multi-task as much. I decided to dedicate each week to a specific project, and just switch main projects every week. Of course, I would still have all my meetings and kept up with collaborations, but when it came to my main project time, the entire week was dedicated to one project. By the end of the week, I would then be able to write a to-do list for myself for the next week I would spend on that project. This "uni-tasking" not only allowed me to check actual big items off my list, which made me actually feel productive, but it also allowed my subconscious processing to kick in during down times / weeks off. ### "*Don't try to do two jobs at once*" My postdoc year, I tried to practice the advice I received from my esteemed CMU colleague Yonatan Bisk, who said "*Don't try to do two jobs at once*," meaning, don't try to do all the professor things and also all the postdoc things. I took that to heart as much as I could, and did not attend faculty meetings at CMU despite being on the mailing lists already. I barely participated in the faculty hiring process that year. I did, however, participate in PhD admissions a bit, as that would be my first time recruiting students for myself. More on that later.... ### Dipping my toes into grant writing For the most part, I did avoid grant writing until roughly February or March, as I was trying to focus on research. *In Feb/March 2022*, I tried to lead writing one small grant proposal myself, for a call by Meta for only $55k (roughly half a student year at CMU). It felt like a good opportunity to learn how to do this, and it didn't take that much time. The proposal was ~2 pages, and I got several rounds of feedback from Yejin to polish it. I ended up getting that money, which was both exciting, and also terrifying as it fed my fear of failure even more (my brain would tell me: "*great job, now that you already know how to do this, you are not allowed to get a grant rejected in the future*"). *In May 2022*, my friend and collaborator Katharina Reinecke, HCI professor at UW, brought me into a grant she was leading on cultural clashes between users and AI. This was a pretty smooth experience (for me at least, probably not for her since she was doing most of the work), and I didn't really have to do much, she basically told me "*here are the places you just need to write some small paragraph outlining a proposed idea*" which I did. Since I didn't have any students yet, I didn't really ask for money to fund anything other than a bit of my summer salary. In retrospect, I probably should have asked for a little bit of CMU student time, but it's fine, it's a great collaboration. Then, *in June 2022*, I was nominated to apply to the AI2050 fellowship, which required an online form to be filled out. That one was a harder one to write, as the premise of the grant was basically "*It's 2050, AI has revolutionized society, what happened?*" which led me to go down many rabbit holes, including one that I occasionally end up in where I think "*AI is screwing up the world, should we really be working on it?*". That aside, I ended up writing something and submitting it. In *summer 2022*, the last grant effort I did was writing my Ai2 Young Investigator Award application, an award that you can apply for if you're a postdoc at Ai2 going on to become a professor somewhere. There was not really much guidance on what to write, how long to write, and who would judge it, but I wrote 2 pages about the same project I proposed for the AI2050 fellowship. I ended up getting it right away basically. ### Taking summer interns I also decided to take two interns the summer that I was working at Ai2, one senior PhD student and one first year PhD student. Having these interns felt a little bit like dipping my toes more formally into the PhD advisor role, for better or for worse. While the senior PhD student did very well and was very independent, the junior intern was very hard to advise. They were quite unfocused, reluctant to commit to a project (despite it being a 12-week internship), and just felt like they didn't want to listen to what I had to say. Taking a slight detour here, but I had at least two experiences where I was mentoring a student and felt like they did not really listen to me. This made me incredibly anxious, as my brain would tell me "*How are you supposed to be a good PhD advisor if you cannot even mentor your master's student or intern?*" This made me realize how hard the advising role really is, and how much room there is for interpersonal issues. Avoiding discussing specifics of their cases, but it was very hard for me as a junior to feel not listened to or not respected; as someone still feeling like they have to prove themselves, it felt confidence-shaking to experience this. Looking back, I do think some of these feelings would not happen now that I feel more confident in my advising abilities. However, at that time, it was very rattling. The best insight I got was from Henny Admoni at CMU, who helped me understand that I'd probably have the capacity to successfully advise *any* PhD student—the (unrealistic?) standard I hold (held?) myself to—but that some students might take a lot more energy than others due to interpersonal issues, and it's okay for me as a junior faculty to want to protect my energy given how hard the job already is. This gave me a little bit of grace to feel less bad about not being able to make it work with everyone. ### Giving talks everywhere! One thing I remember is that after having done the job search, having polished it to death, and having given my job talk in over 8 different institutions, I was sure as hell going to keep giving it! So I definitely said yes to any (remote) talk opportunity that came my way, partly as a way to get my name even more out there, to recruit students, and promote my work. Note that I didn't always give my job talk, sometimes I tailored it more to the venue. ### Recruiting my first PhD students The most faculty-like work I did during my postdoc was recruiting my PhD students. This is to this day still a task I am not confident in my abilities to do "well." This is not a reflection on my students btw, my students are amazing and I am very proud of them! It just feels very hard to make decisions about who to work when there are so many choices, and it's hard to know how to make these choices. *Advertising positions*: I didn't really put out many recruiting calls, other than announcing that I was starting at CMU when I got the job (on Twitter, on my website). When I was at EMNLP, multiple students said they wanted to apply to work with me, and I encouraged them to apply to CMU. I also put up a ["contact me" page](../contact.html) outlining that students should apply to CMU and put my name down. This strategy worked for me as many students applied, but looking back maybe putting out more targeted recruiting calls would have been good as well. *Screening applications*: In my postdoc year, I wasn't actually on the full PhD admissions committee, so by the time I got to see applications, they were narrowed down quite a bit. I chose roughly 8-9 students to do some phone screenings with to see if we had a good fit. Looking back, I had no idea what to ask in phone screens... I ended up relying a lot on some other faculty at CMU and faculty mentors to help me make decisions, and after many many anxious conversations with my mentor. *Making offers*: The advice I got was to make 3 offers, which was the absolute limit of how many students I could support financially— though I didn't expect all 3 students to come but it was a great compliment to me 🥰. Again, I am very happy with my students, and there are many reasons why starting with many was a good thing, but I do think financially that was stressful for me. One of the PhD students I made an offer to was actually a master's student I had worked with at UW before; we had already published a paper together, albeit when we were both more junior. I am happy I recruited this student, because it meant that we were able to get the ball rolling faster on research, we already had a more established rapport, and could even start research a little earlier. While there were some students that I could have co-advised, at that time my research directions were somewhat far from other faculty at CMU, making it harder to find matches for me and another faculty. So I ended up making offers to students by myself. Again, there are advantages and disadvantages to doing this. Solo advising my first students allowed me to completely set the tone and climate of my lab, and allowed me to grow as an adviser on my own; but it does add to the financial burden. If I had co-advised right away, maybe I would have felt less like I had a "lab" of my own, maybe I would have learned advising habits from my co-adviser that didn't feel right for me, and maybe I wouldn't have had the chance to build my own research direction, but I probably would have had less stress financially. *Visit days / open house*: Recruiting PhD students when you yourself have no experience being an advisor can be one of the most stressful aspects of the early faculty life—at least for me it was. Applicants often ask prospective advisors what their ongoing / future research is, what your their advising style is, what their expectations are for students, etc. Since I had not started, I had only tentative answers, which made me feel like an imposter, or a dirty politician asking people to blindly trust them without having any evidence to back them up. I tried to be honest with my prospectives about how I was flexible and planning to grow as an advisor but that I couldn't give definite answers. For some students, I think that might have seemed like I was doubting myself; for others, it might have been a sign of openness and honesty. Looking back, I don't know what the right approach should have been, but since I do feel confident in my advising now, I think I could have been more confident in my advising and research vision (since I had been advising junior students since my later PhD year), while being honest about the things that I did not know how to do yet. *Waiting for the year to start*: After my PhD students accepted their offers, I mostly let them finish their old lives. The only exception is that I did reach out to my one student that I had worked with to help with taking over one of the projects I was trying to lead but didn't have time for. ### Post-Postdoc reflections - Glad I... - Glad I took the advice of not trying to do two jobs at once. Focusing on the postdoc instead of diving headfirst into all the professor duties was the right choice, and gave me the time to breathe, recoup, and relax after the stress of the PhD. This was especially important in January 2022 when my boyfriend of 1.5 years and I broke up, and I had conflict with my mother that ended our relationship forever. This also allowed me to take time to say goodbye to Seattle (my home for 8 years) and all my friends there. - Glad I learned to not multi-task projects and instead try to dedicate one entire week per project. - Glad I was able to kickstart some collaborations and projects (via internships, mentoring students) which turned into publications and made me feel more productive in my first year of faculty. - Glad I "beta tested" some projects that led to workshop papers from mentoring junior students/master/undergrads, which I could draw from when guiding my new PhD students towards a first-year project. - Glad I took the opportunity to work on only a couple of grant proposals early on, which helped me build some confidence and get some "firsts" out of the way, while not being overwhelmed with too much grant writing. - Glad I had my faculty colleagues and mentors to help me through PhD admissions and the anxiety of choosing my first batch of students. Special shoutout to Graham Neubig, and my advisors Noah and Yejin. - Glad I started reading group with CMU junior faculty around An Everyone Culture, even though we didn't really discuss the book much, the conversations really helped me. - Sad that... - Sad that I felt so stressed during PhD visit days / open house. I could have been more confident in my advising and research abilities, even if indeed I hadn't done the full faculty job yet. - Sad that not all my mentoring experiences were successful. I see now that it's okay not to "click" with every student, but at the time, it felt like a personal failure. - Sad that I wasn't able to celebrate my small Meta grant proposal being accepted as much as I wanted to, and instead letting the stress of future failure get to me more than I wish it had. - Sad that I left Seattle. That was my home, and my friends there were my chosen family. Even though I now love my life in Pittsburgh, I still miss them a lot.