LabMate AI

Your Research Co-Pilot

Comment

Inspiration

As students and researchers, we often struggle to navigate vast academic literature, identify what's already been studied, and more importantly — what's missing. I wanted to create a tool that bridges this gap, helping anyone from undergraduates to PhD candidates go from a research question to actionable insights quickly. This inspired me to build LabMate AI.

What it does

LabMate AI helps users:

  • Design experiments
  • Get summarized research on any question
  • Find similar work
  • Identify literature gaps
  • And most importantly, do all of this with academic references.

How we built it

LabMate AI is a full-stack application built using:

- Frontend: Next.js (React)

- Backend: FastAPI

- Database: PostgreSQL

But the real power comes from how the Perplexity Sonar API was integrated, which enables deep academic research capabilities. We crafted custom prompts for four use cases:

- Research Summary

- Similar Work

- Literature Gaps

- Experiment Design

Depending on the task, I used either:

- sonar-deep-research for comprehensive, citation-backed answers

- sonar-reasoning-pro for reasoning-intensive tasks

I fine-tuned how prompts were structured and dynamically adjusted web search context for improved accuracy — especially when finding related work.

Users can save projects and revisit them later through a clean dashboard, making it perfect for iterative research workflows.

Challenges we ran into

  • Prompt Engineering: One of the biggest challenges was crafting prompts that reliably returned high-quality, citation-rich responses. I had to iterate on tone, phrasing, and task-specific keywords to align with Perplexity Sonar's strengths.

  • Handling Diverse API Outputs: Since different tabs (e.g., "Research Summary" vs. "Similar Work") required different models and settings, I needed to dynamically manage model selection and parsing of varied API responses.

Accomplishments that we're proud of

  • Seamlessly integrated Perplexity Sonar API to provide accurate, citation-rich research summaries and suggestions.

  • Built a full-stack system with Next.js, FastAPI, and PostgreSQL, including user authentication and persistent project saving.

  • Created an intuitive UX flow from question input to viewing, saving, and revisiting results.

  • Enabled four core research assistant features: summary, similar work, literature gaps, and experiment design, each supported with real academic sources.

  • Delivered a functional MVP under time constraints while maintaining clean design and reliability.

What we learned

  • How to effectively use the Perplexity Sonar API for deep academic research use cases.

  • The importance of prompt customization to align model output with user expectations.

  • Managing full-stack web application development, including authentication, state persistence, and asynchronous API integration.

  • Structuring a product that supports modular research tasks, while ensuring ease of use and scalability.

What's next for LabMate AI

  • Collaboration Features: Allow users to invite peers or advisors to collaborate on saved research projects.

  • Auto-Citation Tool: Automatically generate formatted citations (APA, MLA, Chicago) from references returned by the AI.

  • PDF Upload & Analysis: Enable users to upload research papers and ask questions directly about their content.

  • Notebook Integration: Add the ability to take notes, tag findings, and organize insights within each project.

Built With

Share this project:

Updates