Learn't

Inspiration

Learn’t is inspired by the need to make education more accessible and equitable for students everywhere. Many students lack access to traditional tutoring services due to financial constraints, busy schedules, or other barriers. To bridge this gap, learn’t offers an adaptable tool designed to help students effectively manage their time, concentrate on areas needing improvement, and minimize stress by providing a structured, flexible study pathway. High schoolers, university students, and lifelong learners alike can benefit from this platform, empowering them to achieve their educational goals with greater ease and focus.

What it does

Learn't customizes a study experience by collecting key information, including the total study time available, preferred session length, chosen topics, and specific learning objectives. With these inputs, Learn't generates PDFs of a tailored study guide, a detailed study planner, and a topic-specific list of practice questions. The study guide provides in-depth coverage of each topic, including definitions, core concepts, testing methods, and example questions. The study planner outlines a strategic study schedule to maximize productivity within the available time, detailing what to focus on in each session, suggested work and break intervals, and recommended activities for review post-session. Additionally, a set of curated practice questions enables learners to apply and reinforce their understanding.

How we built it

The goal of the front-end was to create an effective and visually appealing interface. We used React.js for its component-based architecture to streamline development. Integrating Firebase enabled Google sign-in, allowing users to save their projects easily. We designed a user input form for study goals and preferences, using JSON files for real-time retrieval of personalized study materials. These results are displayed on a well-organized dashboard, where we included a download option and a Gemini chatbot for user interaction. Overall, the front-end is designed to empower learners, making their study experience more personalized and efficient. In developing the back-end, we leveraged Anthropic's Claude API, applying advanced prompt engineering techniques to optimize responses and improve the accuracy of generated content. This required extensive trial and error, particularly in fine-tuning prompts to retrieve data that aligned with user expectations. We also invested time in structuring and formatting the retrieved data, ensuring consistency and clarity across outputs. By iteratively refining the prompts and data handling procedures, we were able to create a back-end system that delivered reliable and precise responses, ultimately enhancing the overall user experience and functionality of the application. The connection between the two ends was essential for a seamless user experience, we used RESTful API calls to enable dynamic application. Data transfer was done through temp JSON files, and API calls from the front-end were connected to a Flask server running on localhost that allowed us to use our backend.

Challenges we ran into

We faced several challenges during development, particularly with the Anthropic API, database management, and frontend-backend connectivity. Managing API keys was initially problematic, which we resolved by using a .dotenv file for secure handling. After debating the necessity of a database for storing generated PDFs, we opted for a JSON file as a lightweight alternative for data exchange. The primary challenge was bridging frontend and backend functionalities, which we addressed by implementing JSON file exchanges to streamline communication.

Accomplishments that we're proud of

We’re proud of several accomplishments, including successfully connecting the front-end and back-end, achieving a polished and user-friendly website display, and seeing the platform run exactly as we envisioned. Additionally, we were able to implement extra features, such as an AI tutor, enhancing the user experience beyond our initial goals. Learning new frameworks and powerful APIs like Anthropic, Firebase, Google Gemini, and Postman was a humbling but extremely enriching experience that we all are proud of.

What we learned

A key strategy we learned was parallel development. Given the tight integration between the front-end and backend of our application, sequential development would have been unproductive. On the backend, we used Postman to work on our AI generators, interactive tutors, and organizers independently of the front-end. Meanwhile, the frontend team developed an organized web application using pre-made JSON structures to simulate API responses. This approach allowed us to efficiently create a functional application that we’re proud of.

What's next for Learn't

The first thing we plan on developing after AI ATL ends is a database through Firebase that will store user accounts, collections of their tasks, and upcoming events, as well as a historical view of past exam planners/study guides. We intend to make the application’s planner aspect into a tracking system that will record user progress and adapt plans based on missed sessions and other user input. Our most ambitious undertaking will be to develop our own natural language processing system that will be able to analyze textbooks, class notes, and other resources to create a knowledge base for our own generative AI system that could utilize popular study techniques like the Pomodoro technique. Our goal is to completely remove the reliance on other NLP systems over time, by uniquely developing our own machine-learning system.