arXiv-Pulse

Inspiration

As a computer science student, I've always been fascinated by the intersection of AI and academic research. The idea for arXivPulse struck me during a late-night study session when I found myself overwhelmed by the sheer number of papers I needed to sift through for my thesis. I
thought, "What if I could create an app that not only helps me find relevant papers but also understands and interacts with them?"

What I Learned

Building arXivPulse was an incredible learning experience. I dove deep into:

1 Next.js and React: I learned how to create a responsive, server-side rendered application.
2 AI Integration: I got hands-on experience with LangFlow and integrated it with Theta's Edgecloud services to aid research paper summarization.
3 Full-Stack Development: From designing the UI to setting up API endpoints, I gained a holistic understanding of web development.
4 Performance Optimization: I learned techniques like infinite scrolling and efficient state management to handle large datasets smoothly.
5 PDF Processing: I explored various libraries and techniques for extracting and rendering PDF content in the browser.

How I Built It

I started with a basic Next.js setup and gradually built up the features:

1 First, I integrated the arXiv API to fetch paper data.
2 I then created a clean, responsive UI using Tailwind CSS, incorporating a very different approach from most academic sites as an experiment.
3 For the AI features, I integrated Langflow powered by models hosted on Theta's EdgeCloud platform, crafting prompts for paper summarization.
4 I added features like citation generation and social sharing to enhance the user experience.
5 Finally, I implemented the PDF viewer and chat interface, allowing users to interact directly with papers.

Challenges I Faced

As a solo developer, I encountered several challenges:

1 Time Management: Balancing this project with my coursework was tough. I had to learn to prioritize tasks and manage my time effectively.
2 Technical Hurdles: Integrating AI was particularly challenging given limited compute availability on the theta platform during the hackathon. This gave me a reason to balance my performance with these constraint using lamgflow to manage and schedule my api calls. While the core summarization feature of the app runs on langflow powered by Theta I had to switch the chat section to Google's Gemini Api to work around performance/availability issues.
3 Performance Issues: As the app grew, I faced performance bottlenecks, especially with PDF rendering and real-time updates. Optimizing these without compromising functionality required a lot of research and experimentation.
4 User Experience Design: Creating an intuitive interface for complex functionality was harder than I anticipated. I had to put myself in the shoes of researchers from various fields to design a universally accessible UI.
5 Scope Creep: I kept thinking of new features to add, which threatened to bloat the project. Learning to define and stick to a core set of features was a valuable lesson in project management.

Despite these challenges, seeing arXivPulse come to life has been incredibly rewarding. It's not just a project for my portfolio; it's a tool I use daily in my studies. The positive feedback from my professors and peers has been encouraging, and I'm excited to continue refining and expanding arXivPulse in the future.