ATLAS

• 

  System architecture of ATLAS showing interactions between the app, AI agent, and database services.

Inspiration

Everyone today wants to learn AI, but most people don’t know where to start. There are too many resources out there, and it’s hard to figure out what’s good or what order to learn things in. We wanted to make something that gives people a clear, guided path to learn step by step.

What it does

ATLAS helps users plan their learning journey. It shows a skill map where they can see how different skills connect. You can chat with the app and say, “I know this, and I want to learn that,” and it will create a personalized learning path for you. You can then follow the resources, track your progress, mark things as complete, and even take notes along the way.

How we built it

The app is built with FastAPI for both the frontend and backend.

• The frontend uses shadcn/ui and HTMX.
• The skill map is visualized using Cytoscape.js.
• PocketBase handles user data and progress tracking.
• KùzuDB is used as a graph database for managing skills, learning nodes, and resources. We deployed everything the app, KùzuDB API service, and PocketBase to AWS EKS. Our AI route planning agent is deployed to AgentCore. When users ask for a learning path, the app talks to the agent, the agent fetches data from the KùzuDB API, and then responds back with a recommended route all through a chat interface. ## Challenges we ran into One major challenge was deploying the agent to AgentCore. The app and the agent were calling each other in a loop like a deadlock. To fix this, we built a separate API service just for KùzuDB. Another challenge was designing a flexible graph schema that could connect all the learning nodes and skills properly. ## Accomplishments that we're proud of We built a working, interactive skill map that actually recommends what to learn next. We designed a clean and modern UI. We deployed our AI route-planning agent to AgentCore. ## What we learned We learned how graph databases can make learning paths more natural and connected compared to regular databases. We also learned how to package and deploy agents to AgentCore, and how to separate services to make them work smoothly together. ## What's next for ATLAS Next, we want to build different skill-level paths so users can choose beginner, intermediate, or advanced routes. We’ll also add team features, where users can create teams, assign roadmaps, and track group progress. And finally, we plan to let ATLAS automatically discover new skills and resources to keep the learning paths fresh and up to date.

Built With

• agentcore
• bedrock
• docker
• eks
• fastapi
• kuzudb
• pocketbase
• python