Inspiration
Everyone advises adopting sustainable habits like using public transportation and recycling to aid the planet. Yet, it often seems like such advice goes in one ear and out the other for many. "Each small step for man, is a giant leap for mankind." With this quote in mind, we created SustainityAI – your personal eco-companion that nudges you daily to take a small step towards helping our Earth.
What it does
The main thing SustainityAI is made for, is to remind us to do one "green" thing a day. They are all simple tasks everybody can do, but things we usually don't. It sends you notifications every morning, and a reminder at evening, to check in on our progress with the task. And no, Sustainity doesn't judge! If you miss a few tasks, that is absolutely fine! The best part? It doesn't just send you random notifications which you can't possibly do. They are always the simplest tasks, personalized for you. When you download Sustainity, you'll be given a short questionnaire that can be filled in 2-5 mins. Using this data, it sends you notifications and reminders that suit you, your need, and your budget! The app itself is not limited to this functionality though. You can chat with Sustainity, and it doesn't need to be related to sustainability issues either. It's basically a free ChatGPT, right at your fingertips. You can also use Sustainity to have fun as well! It uses your information to learn and tell you about fun activities you can do, personalized for you!
How we built it
Tim, responsible with the front-end, used Java to make the desktop app. Simplicity was key, he made sure that the app functions as expected, and notifies you correctly! He is also responsible on making sure the AI runs smoothly on the from end! Taskeen was responsible with the AI, tried and tested multiple variations of Language models, and decided whether to use APIs or to download it directly on the app.
Challenges we ran into
The biggest challenge was integrating the LLM API with the app. Due to limited documentation on Eclipse, we were severely limited to trial-and-error methods to find the best way to use the API, and to send notifications appropriately. Another issue was that since LLMs can take up large amounts of RAM, we had to find the best model which fits the use-case, and also doesn't overload the user's laptop. Due to both the teammates residing in different countries, we had to resort to zoom calls to complete the project.
Accomplishments that we're proud of
We are proud to have an app, complete with chatting functionality and ability to send notifications and reminders. The LLM choice was tough, but we finally stuck to the open-source TinyLlama model, which outperformed other models on a similar scale.
What we learned
We learnt a lot about collaboration, teamwork, and coding. Taskeen learnt a lot about choosing the correct LLM, the parameters he needs to use to accomplish the task, and also a lot about integrating APIs. Tim learned a lot about incorporating the LLM and other AI into the main program, the use of Java to achieve various system functionalities, and the use of data analysis based on user's responses
What's next for SustainaityAI
SustainityAI is an idea which can last. We are planning to add better models once we get the appropriate resources to do so, so that we can make it more personalized. We also plan on improving the latency and call speeds by optimizing our code. We also plan on making a mobile app so that it is more accessible to the public. As Sustainity gains more publicity, we also plan on adding a twist: the bigger streak of completing tasks given by SustainityAI, we can provide FREE OpenAI credits, or other AI coupons to users. This could potentially help them stay motivated to complete the tasks!
Log in or sign up for Devpost to join the conversation.