Anuja Uppuluri used to spend a lot of time scrolling social media apps dictated by algorithms designed to keep users glued to their screens no matter how mind-numbing the content.
“I always had something else that I could be doing or wanted to be doing, but I was choosing to watch TikTok videos for five hours,” said Uppuluri, who completed a bachelor’s degree in information systems at Carnegie Mellon University in May. “And then by the end of it I couldn’t remember anything that I had watched.”
Uppuluri, who now works as a machine learning engineer for Anthropic, sought to become more intentional about the information she consumes and has since scaled back her social media usage. Rather than scroll aimlessly, she wanted to fill her time digesting more research related to her career field, especially about the inner workings and implications of increasingly prevalent generative artificial intelligence tools.
She discovered all types of academic papers, articles and blog posts she wanted to read, but it wasn’t easy to keep organized. “I didn’t know where to put all of this stuff, because there’s no central location for it,” Uppuluri told Inside Higher Ed. “I started thinking about how I want to use my research and what I want to see from other people’s research.”
So Uppuluri developed Paper Trails, which she described as “Goodreads for academic papers” in an X post announcing the website’s launch last week. “I built it because I wanted a place where engaging with research felt fun, beautiful, and personal to you.”
Similar to the book-focused website Goodreads, Paper Trails is designed to help users discover new research and ideas, though it’s not powered by an AI algorithm. It’s a crowd-sourced platform where users can post links to papers from any field, peruse summaries of those papers, create shelves (public reading lists), and comment, review or rate papers.
In the week since its public debut, Paper Trails has grown its users from 10 to 2,200; the number of articles available on the site has increased from 20 to 3,100.
Inside Higher Ed spoke with Uppuluri to hear more about her vision for Paper Trails.
(This interview has been edited for length and clarity.)
Q: What sparked your interest in reading more academic research, especially about computer science and AI?
A: I always thought of a computer scientist as someone who writes code, builds infrastructure and makes sure systems are built to scale. But AI is blurring the line between research and engineering. Every new discovery that comes out of a large language model (LLM) lab is research-oriented.
Understanding why the technology is the way it is can be done by reading papers and understanding the research about large language models. These models are like black boxes—you can’t entirely understand what’s going on inside it—and that’s created all of the research subfields. For example, a subfield called interpretability is about trying to interpret what the models are doing. The more you do with these models, the more you have to read to understand how they work to gain context on how to build things better.
Q: How did your previous experiences reading and writing research papers inform the creation of Paper Trails?
A: A research paper to me used to mean something related to medicine, chemistry or biology. I didn’t understand or realize that it had a place in computer science until a few months before I wrote the only research paper I did in college. For that paper, I wrote the methodology and code, but my co-author, who was doing his Ph.D., helped me structure the paper, write the references and get it through the formatting process. It felt like a high barrier to entry for doing and reading research, which I associated with work instead of fun.
I thought having some type of casual thing like [Paper Trails] to organize research papers would maybe help me read more. But other existing websites I experimented with looked so dated and used software I had to learn. It looked complicated and didn’t seem fun. I also didn’t want to organize my research in a big Google Doc that has 50 links on it that I’m never going to touch again—that looks ugly.
So, I knew I could probably make my own site that looks nice and is easy to use.
At its core, Paper Trails is a tool to put papers and other reading material together in a way that’s pretty and fun. Sometimes that’s what you need to make something feel more like a hobby rather than more work you have to do.
Q: What went into developing the Paper Trails website and how does it work?
A: I coded most of it from scratch, with the exception of pasting in a few codes to fix some bugs.
When I first launched, there were only around 20 papers on the site. Now, there’s around 3,000 just from more people being on the site and adding the things that they want to read.
I chose not to mass import a bunch of stuff at the outset because if people look at it and it’s not something they’re interested in, it’s still there. It’s kind of cool to look at every single paper and know that it’s there because it’s something on someone else’s reading list.
Everything on the main page is organized by publication date. You can also use keywords to search or just click some buttons to see what people are logging. There is no personalized algorithm for users. While there isn’t anything necessarily wrong with a recommendation assistant, it’s also nice when there’s nothing telling you what to look at.
Q: What is the value of Paper Trails for its users?
A: There’s a lot of people who would like to get into research or just reading more. And if you want to spend your time in that way, having a tool to help you do it and encourage you to do makes it a lot easier to follow through on.
There’s also value coming from all the people that make it a collaborative thing. It allows people to explore, kind of like going down a Wikipedia rabbit hole. You can just keep clicking on random links and reading. You don’t know what you’re going to learn at the start of your session but, if you’re interested in it, you can read it.
Q: Now that people are showing interest, what are the next steps for Paper Trails?
A: I was thinking about sending it to some of my old professors, especially if they have Ph.D. students who may be interested in working on it.
There are even more elements that I could add that would improve the user experience. A lot of people have papers that are already saved on another site, so being able to bulk import would be helpful. Or allowing a few people to edit a shelf rather than just one person could make collaboration a little bit better. Or being able to clone somebody’s shelf so that another user can add some of their own stuff to it.
I don’t know exactly what growth looks like. But to me, success means the people using it are happy.
