TimeWiseAI

TimeWiseAI

Inspiration

As a High School student myself, I found myself having trouble juggling all of my extracurriculars and classes. I often had time conflicts and wasn’t able to completely optimize my time. I also observed many of my peers and classmates facing this exact issue. To fix this, I created a spreadsheet with my entire week’s schedule. However, even though this format wasn’t ideal, it gave me an idea for a bigger solution. I decided to utilize my extensive knowledge and experience in AI and Programming to build a fully-functioning app which is designed to neatly organize and manage a student’s time. I decided to incorporate a Large Language Model (LLM) in my app to create a customized chatbot which significantly lightens stress amongst high schoolers by providing a guide full of advice and tips. This includes an estimate of the ideal number of hours to aim for (per activity/category), pointers on how a student can optimize their performance inside and outside of school, healthy habits backed by science to increase productivity, etc. My own personal experiences, coupled with issues I noticed in my community and age group, inspired me to create an efficient time-management and academic productivity tool: TimeWiseAI.

What it does

TimeWiseAI is an efficient time optimization tool designed for students. The purpose of TimeWiseAI is to help students easily manage their academics, extracurriculars, and other activities through an AI-based interface that provides statistics on the user’s performance in targeted areas, uses a neat notification system to keep the student on track and gather feedback on their performance, and suggests a plan for organizing their activities.

The main feature of this app is a personalized chatbot that uses the data from the student’s entire history tasks and their performance to provide customized advice, comprehensive guides, and solutions to help the student organize all of their events. The chatbot is integrated with OpenAI’s ChatGPT to create a LLM (Large Language Model) which uses NLP (Natural Language Processing) after the user enters a brief prompt and requests the TimeWiseAI chatbot to suggest the most optimal practices for maximum success, comfort, and mental wellbeing.

My app also includes a built-in calendar generator and interface. This consists of 3 calendar screens: a Daily calendar, Weekly calendar, and Monthly calendar. This allows the student to organize all of their classes/homework. There is an AI-powered event generator which provides a fully customized calendar for an entire week, month or year, based on the user’s short request. This process happens in under a minute! The user can optimally use a voice-to-text feature to make the event generation easier. Also, the TimeWiseAI understands all languages, and this multilingual flexibility offers a lot of advantages to students worldwide.

How we built it

When coding TimeWiseAI, some of the issues I faced were minor bugs in the code, while others were relatively large. When I encountered larger problems such as network errors or backend issues with the GPT-4 model’s async functions, it usually took several hours to fix because of difficulties with connecting to the remote environment of the OpenAI API. Adjusting the temperature (randomness of the response) was also quite tricky. This is mainly due to the fact that AI is evolving very rapidly, and the API performance and response varied over my period of software development. I started with using GPT-3.0, moved to GPT-3.5 AI models and am finally using the state-of-the-art GPT-4 model with the app

Another large difficulty I faced was issues with my package manager. In order to solve this, I had to dig deep and identify the root causes of my problem. After an intensive session of correcting all of the issues with the packages and dependencies, I was finally able to finish the feature which allowed for efficient navigation between the screens.

I have also faced a few bugs that were relatively minor to the bigger issues, but they were still technical difficulties nonetheless. For example, when I began coding the app, my component tree/structure was not ideal. This made it difficult to facilitate object inheritance. Once I re-created my entire file infrastructure, I was able to easily program newer components into my app. Another small issue I faced was related to the user’s events list. Analyzing the output of my AI Model and modifying it to fit the format of the JavaScript array required me to create a parsing feature which included an intensive process. This created several minor bugs, but I was able to fix them after deep sessions of debugging and refreshing/rebooting.