AutoPeer

Inspiration

We were inspired by an astrophysicist who was deeply interested in how AI will change the way scientists conduct research and experiments. This inspired us to think about how AI agents could be a great first step in providing critiques that help authors better prepare their papers for submission to journals.

What it does

AutoPeer takes a scientific paper and writes a professional critique. It doesn’t replace human peer reviewers but offers authors a quick and easy way to gain feedback and brainstorm with the AI on how to improve their paper before submitting it to a journal.

How we built it

We built AutoPeer using Streamlit as the front end and OpenAI Agents, with each agent assigned a specific task to analyze and critique the paper.

Challenges we ran into

Initially, the PDF reader wasn’t translating the data correctly for the AI to use. However, we managed to overcome this challenge by restructuring the agents to work more effectively with the PDF reader.

Accomplishments that we're proud of

We’re proud to have created a tool that can be used in day-to-day research. AutoPeer makes a scientist’s job easier by providing quick feedback, saving time, and reducing the wait for a traditional peer review, which often takes months.

What we learned

In addition to learning how to build applications with AI agents, we also gained valuable insights into the research process and how AI can significantly enhance certain aspects of it.

What's next for AutoPeer

In the future, AutoPeer could become a one-stop shop for researchers, fully automating the peer review process and even assisting with scientific research and complex computations. This would allow scientists to focus on creating new theories and discussing ideas, while leaving the repetitive tasks to AI.

Built With

• faiss-cpu
• langchain
• langchain-community
• openai
• pypdf2
• python
• streamlit
• tiktoken