๐ฏ Inspiration
Learning online can be overwhelming. Resources are scattered, unstructured, and often too generic or behind paywalls. I wanted to create a tool that could understand a learnerโs unique constraintsโlike time, budget, and content preferenceโand instantly generate a customized, actionable roadmap. Thatโs how CourseCrafter was born.
โ๏ธ What it does
CourseCrafter generates personalized learning plans using an AI language model.
Users input:
- a topic,
- how much time they have,
- their budget,
- and preferred content type (e.g., video, text, etc.)
The app then returns a detailed, well-structured course plan with real resource links and a study schedule. The user can also download the plan as a .txt file.
๐งฑ How I built it
- Frontend: I used Gradio to build a clean and interactive UI.
- Backend: I deployed an AWS Lambda function to parse user input and forward it to the LLM.
- AI Engine: I used Modal to host a FastAPI server running a large language model (Meta LLaMA 3 or Mistral).
- Deployment: The Gradio app runs on Render or Hugging Face Spaces, while the backend and model are served through AWS Lambda and Modal.
๐ง Challenges I ran into
- ๐ Debugging nested JSON payloads in AWS Lambda
- โ Handling CORS errors and HTTP 500s from API Gateway
- ๐พ Managing Git LFS and "no space left on device" errors in Gitpod
- ๐ง Learning the new Modal SDK structure and deploying GPU-backed models
- ๐งช Testing and validating the end-to-end data flow between Gradio, Lambda, and the LLM
๐ Accomplishments that I'm proud of
- Successfully built and deployed a full-stack AI-powered application end-to-end
- Integrated multiple services (Gradio, AWS, Modal) seamlessly
- Learned how to deploy and optimize inference for large language models
- Created a tool that can actually help learners save time and get started quickly
๐ What I learned
- Deploying serverless APIs with AWS Lambda + API Gateway
- Hosting and calling LLMs using Modal + FastAPI
- Building responsive and user-friendly interfaces using Gradio
- Troubleshooting deployment, file size, and data parsing issues
- Thinking holistically about how AI can provide value in real-world tools
Built With
- amazon-web-services
- fastapi
- gradio
- lambda
- python