Inspiration
Research is everywhere, but finding reliable information and understanding conflicting evidence can be difficult. We wanted to build something that goes beyond simply asking AI a question. This inspired us to create Intellex, an AI system that approaches research like an investigation—examining evidence, contradictions, limitations, and gaps before reaching a conclusion.
What it does
Intellex is an Agentic AI Research Intelligence System that investigates a user's question through multiple specialized agents.
It breaks down questions, analyzes supporting and opposing evidence, identifies contradictions and limitations, highlights research gaps, and generates a confidence score and overall verdict. The results are presented through an interactive dashboard with dedicated sections for Overview, Evidence, Contradictions, and Research Gaps.
How we built it
We built Intellex using React and Vite for the frontend and Python Flask for the backend, with Google Gemini powering the AI investigation.
The system follows a four-agent workflow:
Research Agent → Evidence Agent → Contradiction Agent → Insight Agent
The agents work together to analyze a question and return structured research insights that are displayed in the Intellex dashboard.
Challenges we ran into
One of our biggest challenges was making the AI respond specifically to the user's question instead of generating generic information.
We also faced challenges with connecting the React frontend to the Flask backend, integrating the Gemini API, structuring AI responses into reliable JSON, and designing an interface that could present complex research analysis clearly.
Accomplishments that we're proud of
We are proud to have built a functional multi-agent AI research system rather than a traditional chatbot.
Our key accomplishments include:
Multi-agent research workflow Evidence analysis Contradiction detection Confidence-based findings Research-gap identification Interactive research dashboard Gemini API integration Fallback demo mode for reliable demonstrations
What we learned
We learned that building an effective AI application is not just about connecting an API. The reasoning workflow, prompt design, structured outputs, and user experience are equally important.
We also gained hands-on experience with React, Flask, API integration, structured AI responses, and designing applications around agentic AI workflows.
What's next for Intellex
We plan to make Intellex more powerful by adding real-time web research, verified citations, source credibility analysis, deeper multi-agent collaboration, knowledge graphs, research timelines, and exportable research reports.
Our long-term goal is to turn Intellex into a research co-pilot that doesn't just answer questions—it investigates them.
Log in or sign up for Devpost to join the conversation.