• 

ResearchLens

From papers to research opportunities.

Inspiration

Researching a topic often means reading multiple papers, comparing their approaches, identifying what is missing, and figuring out what could be explored next. For students and early-stage researchers, this process can be time-consuming and difficult to organize.

We wanted to build something that goes beyond simply summarizing papers. ResearchLens was created to help turn existing literature into a starting point for new research.

What it does

ResearchLens is an AI-powered research assistant that helps users move from literature review to research direction.

Users can upload up to five research papers and provide a research topic. ResearchLens then:

• Analyzes each paper and extracts its research problem, methodology, datasets, findings, metrics, and limitations.
• Creates a research landscape showing common themes and different approaches.
• Identifies research gaps across the uploaded papers.
• Detects contradictions and differences between findings.
• Generates potential research opportunities based on the identified gaps.
• Provides supporting evidence from the analyzed papers.
• Lets users Challenge Their Idea by checking their proposed research direction against the uploaded literature and suggesting ways to differentiate it.

The core workflow is:

Papers → Analysis → Landscape → Gaps → Contradictions → Opportunities → Your Idea

How we built it

ResearchLens was built as a full-stack web application.

Frontend

• React
• Vite
• Tailwind CSS
• Lucide React

The interface was designed as an editorial-style research workspace rather than a traditional dashboard, with a focus on readable typography, evidence, and minimal visual clutter.

Backend

• Node.js
• Express.js
• Multer for PDF uploads
• pdf-parse for extracting text from research papers

AI

We use the Gemini API to analyze the extracted research text and generate structured outputs for the research landscape, gaps, contradictions, opportunities, and idea evaluation.

Deployment

• Vercel for the frontend
• Render for the backend
• Environment variables are used to keep the Gemini API key server-side.

Challenges we ran into

One of our biggest challenges was making the AI pipeline reliable enough for a live demonstration.

Our initial implementation used multiple AI requests for individual paper analysis, landscape synthesis, gap detection, contradiction detection, and opportunity generation. This resulted in rate-limit problems and unreliable analysis during testing.

We redesigned the pipeline to consolidate the major research analysis into a single structured AI request.

We also had to solve several deployment-specific issues, including:

• Connecting the Vercel frontend to the Render backend.
• Configuring environment variables correctly.
• Handling PDF extraction on the deployed backend.
• Updating the AI model as model availability changed.
• Making sure the frontend response structure matched the backend's new combined response.
• Fixing API calls that were still pointing to localhost after deployment.

Accomplishments that we're proud of

We are proud that ResearchLens goes beyond the typical "upload a paper → get a summary" workflow.

The project connects several stages of research discovery into one experience:

Understand → Compare → Question → Discover → Validate

We are particularly proud of:

• Building a functional end-to-end research analysis pipeline.
• Turning multiple papers into a structured research landscape.
• Connecting identified gaps to concrete research opportunities.
• Including evidence from the uploaded literature rather than presenting completely unsupported AI-generated ideas.
• Adding Challenge My Idea, which encourages users to critically evaluate their own research direction.
• Deploying the application with a separate frontend and backend.
• Designing the interface around clarity and readability rather than a generic AI dashboard.

What we learned

Building ResearchLens taught us that creating an AI application is not just about connecting an LLM to a prompt.

We learned the importance of:

• Designing structured AI outputs that the frontend can reliably consume.
• Managing API limits and model availability.
• Keeping API credentials secure on the backend.
• Designing prompts around evidence rather than generic generation.
• Handling failures gracefully in a user-facing application.
• Making frontend and backend data structures consistent.
• Testing the deployed application rather than relying only on local development.
• Designing AI features around an actual user workflow instead of adding AI simply for the sake of using AI.

Most importantly, we learned that the valuable part of an AI research tool is not just generating information, but helping users reason about that information and decide what to investigate next.

What's next for research-lens

ResearchLens is currently focused on analyzing a user's own collection of papers. Future versions could expand this into a much larger research discovery platform.

Potential next steps include:

• Automatic paper discovery from academic sources.
• Citation and reference graph visualization.
• Larger literature collections.
• Better evidence tracing and source verification.
• Semantic search across uploaded papers.
• Timeline views showing how research in a topic has evolved.
• More detailed comparison of methodologies and datasets.
• Exporting literature reviews and research-gap reports.
• Personalized research-roadmap generation.
• Integration with academic databases and reference managers.

The long-term goal is to make ResearchLens a tool that helps students and researchers move through the entire journey:

Find literature → Understand it → Compare it → Find what is missing → Form an idea → Test the idea → Start researching.

Built With

• api
• express.js
• gemini
• lucide
• multer
• node.js
• react
• render
• tailwind
• vercel
• vite