The P in Precog Stands For People
I usually don’t remember a lot of things, but what PK said to me on my first day at Precog has stuck with me. “I am on my own journey, it’s your decision whether you want to join me or not”. It might not seem a lot, but to me, that was a challenge. Can I really maximise this opportunity I have been given?
Two A* first-author papers, two workshop papers and one best paper award later, as I prepare to join the University of Cambridge, I can confidently say I maximised my opportunity to the limit. Throughout my two years at Precog, I learnt many things, but the most important was how much an exceptional environment shapes you as a researcher and as a person. Precog gave me exactly that.
The Prologue
As soon as I arrived at IIIT Hyderabad, I told people that I was clear in my goal of pursuing an MS in the US to get a nice cushy job. Back then, I was not sure of what I wanted to do, so it seemed like a safe bet. My seniors immediately pointed me to Precog & PK. I quickly got to know that getting into Precog significantly increased your chances of getting a seat in those places—in some ways, PK’s LoR was talked about like a golden ticket. As soon as the Precog recruitment cycle started, I gave up everything and spent the winter break working on the application task. Around the same time, I was getting very interested in network analysis and graph neural networks. Fortunately, Arvindh was working on the exact same topics in his research. Discussing my interest with him solidified my decision of joining the lab. Little did I know that Arvindh and I will go on to work together on many projects later—long after he graduated. My 30+ hours of working on my application bore fruit as I was formally inducted into the lab.
During the interview with PK, I had mentioned that one of the reasons I wanted to join precog was the freedom that the lab provides. At the start of my stint with precog, this very freedom proved to be something I struggled to handle. As per the norm, newbies are paired up with more experienced students on their projects so they can get a soft launch into their research careers. However, I took a very different route to get to where I am right now. My first project was not even a research project—It was a collaboration to deliver a community detection software. My first semester of working at Precog felt shaky. I had no good research ideas and was just floating by in the lab, thinking hard about whether I’m really built for research. Back then, I had a lot of the required qualities (patience, resilience, etc), but one skill was holding me back—collaboration. And as it turned out, collaboration trumps all your individual skills.
Switching to 7th Gear
I realised if I want to achieve my goals, I needed to be proactive. So, I started reaching out. At the start of 3-2, in January, I took an Independent Study with Prof. Charu Sharma to work on the intersection of Higher Order Topology and AI Explainability. Thanks to my earlier realisation about collaborations, I decided not to do it alone and brought along Sreeram for the ride. Working as team made me realise quickly that a huge part of research is throwing out ideas and shooting them down as a team, which is where Precog’s famous WUs came into the picture.
Sreeram and I did not find much success initially in our search of a “good” research problem. It wasn’t until the last week of March in that semester that we struck gold. And for me, I struck gold twice. I had two great ideas on the same day. The first idea culminated in the eventual work Sreeram and I did, and was accepted at AAAI’25 and our project helped us build an intellectual friendship where we would ask for each other’s opinions and advice on projects. He also recently wrote about his Precog experience.
The other idea I had turned out to be my most valuable experience with Precog. I pitched this idea of Corrective Unlearning on GNNs to Arvindh and Shashwat Goel, who were immediately on board even though it was their last few months in the lab. Working with them developed my research taste and aptitude to a whole new level. I was ready to incorporate all of my learning into this project.
The Final Stretch
As we neared the start of a fresh academic year, the next round of Precog inductions took place. We immediately onboarded Debangan and Sumit onto the Corrective Unlearning project, recognising their potential and interest. This project was a particularly interesting affair as I was receiving mentorship and also simultaneously providing it. We eventually submitted this work to ICLR’25 with high hopes, but were borderline rejected. By this time, I was in my final semester at IIIT-H and was done with all my university applications, so there was no real incentive for me to work hard and resubmit this work to some other conference, right? Even so, I pushed hard for it to be submitted to ICML, whose deadline was just 7 days after our rejection. Needless to say, a lot of work and sleepless nights went into somehow submitting it on time. During this time I realised that research really excites me a lot, and I want to do nothing else (for now at least!).
And so, even in my final semester, I started a new collaboration with Prof. Chirag Agarwal from University of Virginia, on the intersection of GNNs and LLMs. This time, I took a backseat and served more as an advisor/mentor for the students working on the project (Soham, Hari, Anirudh). During this time, I became more involved with administrative lab duties (I was already managing the website since 3-1), and was heavily involved in recruiting the next batch of Precoggers. It felt like I had come full-circle when I sent out acceptance mails to the new batch—exactly like the one I had received two years ago.
4-2 ended with a bang. Arvindh, Karuna, Bala and I got a Best Paper Award for our collaboration with Deepmind, and our work on Corrective Unlearning was accepted at ICML’25, marking my 2nd first author paper at an A* conference. And most importantly, I received an offer to study Machine Learning at the University of Cambridge.
Maybe The Real Precog Was The Friends We Made Along The Way
Throughout my two years at precog, I learnt many things, but the most important was how much an exceptional environment shapes you as a researcher and as a person. I am particularly thankful to all my collaborators, but also to all the friends I made in the lab who helped me grow intellectually and as a person. Kshitijaa, Shashwat Singh, Vamshi, Priyanshul always gave me solid research and life advice, and made the lab a very welcoming place. Thanks to Raghav, Ishwar and all my juniors for always making lab a very fun place to be. And lastly, I am super grateful to PK for believing in me and letting me have the freedom I needed to flourish.
I usually don’t remember a lot of things, but I will always remember the impact that the Precog community made on my research journey. And hopefully, I made a small impact in the lab’s journey forward as well.
You May Also Like
