StudyPlanner AI

I was inspired by the endless cycle of “I’ll study more tomorrow” that many students (including myself) face. Traditional planners are static, while AI‑driven assistants can adapt to a learner’s progress, workload, and motivation.

During this hackathon I learned how to:

Design a full‑stack, serverless architecture – a static glass‑morphic SPA hosted on Vercel talking to a FastAPI backend deployed on AWS ECS. Integrate with Amazon Aurora PostgreSQL – using SQLAlchemy & psycopg2‑binary for secure, low‑latency data access. Implement JWT‑based authentication and subscription monetization – adding /upgrade and /subscription endpoints that gate premium features. Leverage large‑language‑model prompts – the planner generates weekly study schedules with concise LaTeX‑rendered formulas such as [ \text{Study}_{\text{next}} = \frac{\text{Goal} - \text{Progress}}{\text{Weeks remaining}} ]

Deploy a production‑grade CI/CD pipeline – automatic pushes to GitHub trigger Vercel and ECS deployments. Challenges overcome

Cross‑region latency – tuned Aurora read‑replicas and added connection pooling to keep API response times < 200 ms. Dynamic UI on a static host – built a SPA with vanilla JavaScript that loads the backend URL from environment variables, enabling instant feature toggles without recompiling. Secure token handling – stored secrets in AWS Parameter Store and Vercel’s environment variables, never committing them to the repo.