PathAI

PathAI instantly turns scattered web resources like YouTube videos into personalized, vetted learning roadmaps. Stop wasting hours planning—just scrape, sequence, and start building.

Comment

PathAI 🚀

💡 Inspiration

The internet is a goldmine of information, but for most students, it's also a maze of "tutorial hell." We wanted to eliminate the hours wasted on "planning to learn" and replace them with a system that creates a direct line from curiosity to mastery.

🤖 What it does

PathAI is an AI-powered orchestrator that instantly generates personalized, vetted learning roadmaps. It doesn't just provide links; it scrapes, validates, and sequences the best free resources on the web, allowing users to spend less time searching and more time building.

🛠️ How we built it

We designed a decoupled microservices architecture to ensure high performance at zero cost:

  • Frontend: Built with React, Tailwind CSS, shadcn/ui, Lucide React, and Framer Motion for a smooth, professional UI.
  • State Management: Utilized Zustand for efficient, feature-based state handling.
  • Database: Integrated Google Firestore for real-time roadmap persistence and user progress tracking.
  • AI Backend: Developed custom Python scrapers and "Scouts" hosted on FastAPI (Hugging Face Spaces) to handle roadmap generation and video validation.
  • DevOps: Established a fully automated CI/CD pipeline using GitHub Actions and Firebase Hosting.

✨ Gemini 3.0 Integration

PathAI is powered by Google’s Gemini 3.0 Flash, utilizing its advanced reasoning to drive two core agentic services:

  • Curriculum Architect: Leverages Gemini’s superior logic to decompose complex goals into dependency-aware JSON learning paths tailored to specific difficulty levels.
  • Intelligent Resource Validator: Critically analyzes YouTube metadata to filter out irrelevant content via semantic context and selects the highest-quality resource with a natural language justification.

🚧 Challenges we ran into

The main hurdle was bridging the "Body" (Firebase) and "Brain" (Hugging Face). We solved complex CORS issues, managed GitHub Secret injections, and resolved 403 IAM permissions. We also implemented an ethical scraping layer that respects YouTube platform guidelines while maximizing data quality.

🏆 Accomplishments that we're proud of

We are incredibly proud of shipping a full-stack project from start to finish in just 36 hours. Successfully connecting a decentralized system—where the UI, database, and AI logic live on different platforms—was a major technical milestone for our team.

🧠 What we learned

We learned the true power of automated DevOps. Transitioning from manual deployments to a push-to-deploy workflow taught us how professional software is built and secured. We also gained deep insights into Microservices and how to handle environment variables in a CI/CD pipeline. Also, that we make a good team! 😜

🚀 What's next for PathAI

Our next steps involve refining the Video Scouter logic to include community ratings and adding a "Collaborative Roadmaps" feature where users can share their AI-generated paths. We also plan to integrate more diverse content sources like research papers and interactive playgrounds.

Thank you for viewing our project!

Built With

Share this project:

Updates