Cognos AI assitant

Inspiration

Learning with AI today is fast, but fragile. We realized that while tools can instantly explain any concept, users often forget just as quickly because the process of understanding is lost. Traditional note-taking captures final answers, not the journey of how those answers were formed. We were inspired to rethink this entirely: what if we could preserve not just knowledge, but the path of reasoning that led to it? This idea became the foundation of Cogniflow—an AI system that remembers how you learn, not just what you learn.

What it does

Cogniflow is an AI-powered Second Brain that transforms every learning interaction into a structured, replayable learning flow. As users chat with the AI, the system extracts key concepts and builds a dynamic knowledge graph representing their understanding. It tracks how new ideas connect to existing knowledge, capturing the step-by-step journey of learning—including intermediate concepts and reasoning paths. Users can revisit any concept and replay how they originally understood it, reinforcing memory through context. Beyond that, Cogniflow generates AI-suggested alternative learning paths, allowing users to explore new ways of understanding concepts based on their knowledge graph. The result is a system that adapts explanations to the user, reveals knowledge gaps, and evolves alongside their thinking.

How we built it

We designed Cogniflow around three core components. First, an AI-driven structuring layer processes raw user interactions and extracts concepts and relationships, forming a knowledge graph. Second, we implemented a learning flow engine that records each session as a sequence of understanding steps, linking new concepts to the user’s existing knowledge base. Third, we built a reasoning layer that retrieves relevant nodes and traverses the graph to generate personalized explanations. On the frontend, we created a multi-panel interface combining a chat system, a learning timeline, and a graph visualization, allowing users to see their knowledge evolve in real time. The system also includes a path-generation mechanism that uses graph traversal to suggest alternative learning flows.

Challenges we ran into

One of the biggest challenges was representing “understanding,” which is inherently abstract, in a structured and computable way. Deciding how to model concepts, relationships, and learning flows without overcomplicating the system required multiple iterations. Another challenge was bridging the gap between AI-generated responses and a persistent knowledge structure—ensuring that each interaction meaningfully updated the graph. We also faced UX challenges in making complex ideas like knowledge graphs and learning flows intuitive and not overwhelming for users. Finally, generating meaningful alternative learning paths without producing irrelevant or confusing suggestions required careful design of graph traversal and filtering logic.

Accomplishments that we're proud of

We successfully built a system that goes beyond traditional AI chat by capturing and visualizing the process of learning. One of our proudest achievements is the ability to replay learning flows, allowing users to revisit how they understood a concept rather than just re-reading it. We also implemented AI-generated alternative learning paths, enabling the system to suggest new, valid ways of understanding concepts based on existing knowledge. Most importantly, we created a working prototype that demonstrates a shift from static knowledge storage to dynamic, evolving understanding.

What we learned

We learned that effective learning is not about information access, but about connection and context. Building this project showed us how important it is to model knowledge as a network rather than isolated pieces. We also gained insight into how AI can move beyond answering questions to actively guiding understanding. From a technical perspective, we learned how to combine structured data (graphs) with unstructured AI reasoning to create more meaningful interactions. Additionally, we realized that UX plays a critical role in making complex systems approachable—how information is presented can be just as important as the intelligence behind it.

What's next for Cognos AI assitant

Next, we want to improve how the system understands and adapts to the user’s knowledge level in real time, making explanations even more personalized and precise. We also plan to enhance the knowledge graph with better visualization and interaction tools, allowing users to explore and edit their understanding more naturally. Another key direction is refining the generation of alternative learning paths to make them more context-aware and insightful. In the long term, we envision Cogniflow becoming a fully integrated learning companion that can support multiple domains, work offline with optimized models, and continuously evolve with the user—serving as a true extension of their thinking process.

Built With

• amazon-web-services
• and-adaptive-explanation-generation-typescript-+-react-?-frontend-interface-for-interactive-chat
• and-generating-both-real-and-alternative-learning-flows-openai-api-?-for-concept-extraction
• and-knowledge-graph-visualization-node.js-?-api-gateway-to-handle-communication-between-frontend-and-ai-backend-aws-(ec2-/-lambda-/-s3)-?-cloud-infrastructure-for-deployment
• and-learning-flows-for-rapid-prototyping-react-flow-/-d3.js-?-visualization-of-the-evolving-knowledge-graph-websockets-?-real-time-synchronization-between-chat
• and-storage-json-based-graph-storage-?-lightweight-storage-of-concepts
• compute
• contextual-understanding
• d3.js
• ec2
• graph
• json
• lambda
• langgraph
• learning-timeline
• managing-reasoning-paths
• node.js
• openai
• python
• react
• reactflow
• relationships
• s3
• typescript
• websockets