I used the following email to land an undergraduate research position. I found it difficult to contact professors and get started with research at UCSD. Especially at a public school, it is extremely difficult to get noticed by professors.
First some background. I lost two years of university to Covid, which prevented me from meeting professors and getting involved with campus life. I had been interested in working on teaching computers to understand language for several years. But, I got rejected by the two main natural language processing professors after talking to them. One told me to take his graduate class first. Taking his class would have taken me a year to wait for the next offering and then take it without even being enrolled in the class, due to its popularity.
A few months later, I randomly came across a new junior professor who I decided to email; I got an acceptance email 17 minutes later.
I spent a lot of time1 tweaking this email, and I would recommend taking it and modifying it for your needs. If this helps you, I’d love to know! I still believe cold emailing is the best method as it is too difficult to get in touch with established faculty at a large school. Reaching out through public office hours or through a grad student can work, but I have never heard of this working. However, I have seen people get onto projects by asking at the end of classes they did well in.
Hello Professor Elsherief,
I hope you are having a good summer.
My name is Samir Rashid and I am a third year Math-CS major at UCSD. I read your latent hatred benchmark paper and am intrigued by the research you are doing. I would really appreciate an opportunity to get involved with your research if you are still looking for undergraduates.
I am interested in the effects these social media models are having on society. I saw the daunting task of social media moderation firsthand when I interned at Twitter last year. This area is a hugely unsolved issue, and I can see how current limitations are causing huge issues in the world.
I finished CSE 156 with Nakashole last quarter and loved the content. Since finishing 156 I have been following Berg’s 291 to get more practice with NLP topics and to get a better grasp on using PyTorch. I really liked the projects from 156, and the 291 assignments are allowing me to see how to approach more open ended goals. I would love to gain research experience because I cannot fully grasp topics from reading papers alone.
It is my ambition to pursue a PhD degree in CS studying NLP. I have always been interested in language as I studied Latin for 6 years in school and made a program to read Latin poetry in the correct rhythm using Google TTS and some custom audio processing. During the last few years I have seen the acceleration of advancements in language modelling, which has made me want to join the frontiers of language modelling even more. Your research matches perfectly with my interests. I think I am uniquely qualified to work in your lab as evidenced from my course rigor (attached) and my internship at Twitter.
I am able to work full-time over summers and up to 20 hours/wk during the quarter. I just finished my second year and if given the opportunity I plan to commit 2+ years to this position depending on if I continue with the masters program here. I have experience working with robotics with the Unmanned Aerial Systems (UAS) club at UCSD. I worked on making visualizations and writing path planning code for an autonomous plane. I have debugged everything from hardware issues to docker networking to profiling our code to experimenting with different algorithms to see what works best. We ranked 5th at the international competition last month (#1 among US universities). Previously I’ve also worked on building an end-to-end Monte-Carlo localization system for autonomous driving. I have attached my resume and transcript. I would love to speak with you. I am free any day except Tuesday.
I look forward to hearing from you!
This post is certified® human written. I wrote this email before ChatGPT was launched. ↩