AtlasAI | Devpost

Calendar Homepage
Schedule Recommendations
Stress Graph
Activity Recommendations
Customized Spotify Playlist
TogetherAI fine-tuned Model
Backend route to generate new recommended schedule

Inspiration

In today's fast-paced world, highly driven individuals often overwork themselves without regard for how it impacts their health, only experiencing the consequences when it is too late. AtlasAI aims to bring attention to these health issues at an early stage, such that our users are empowered to live their best lives in a way that does not negatively impact their health.

What it does

We realized that there exists a gap between today's abundance of wearable health data and meaningful, individualized solutions which users can implement. For example, many smart watches today are saturated with metrics such as sleep scores and heart rate variability, many of which actually mean nothing to their users in practice. Therefore, AtlasAI aims to bridge this gap to finally empower our users to use this health data to enhance the quality of their lives.

Using our users' individual health data, AtlasAI is able to:

suggest event rescheduling
provide targeted, actionable feedback
recommend Spotify playlists depending on user mood

How we built it

Our frontend was built with NextJS, with styling from Tailwind and MaterialUI.

Our backend was built with Convex, which integrates technologies from TerraAPI, TogetherAI and SpotifyAPI.

We used a two-phase approach to fine-tune our model. First, we utilized TogetherAI's base models to generate test data (a list of rescheduled JSON event objects for the day). Then, we picked logically sound examples to fine-tune our model.

Challenges we ran into

In the beginning, our progress was extremely slow as AtlasAI integrates so many new technologies. We only had prior experience with NextJS, Tailwind and MaterialUI, which essentially meant that we had to learn how to create our entire backend from scratch.

AtlasAI also went through many integrations throughout this weekend as we strove to provide the best recommendations for our users. This involved long hours spent in fine-tuning our TogetherAI models and testing out features until we were satisfied with our product.

Accomplishments that we're proud of

We are extremely proud that we managed to integrate so many new technologies into AtlasAI over the course of three short days.

What we learned

In the development realm, we successfully mastered the integration of several valuable third-party applications such as Convex and TogetherAI. This expertise significantly accelerated our ability to construct lightweight prototypes that accurately embody our vision. Furthermore, we honed our collaborative skills through engaging in sprint cycles and employing agile methodologies, which collectively enhanced our efficiency and expedited our workflow.

What's next for AtlasAI

Research indicates that health data can reveal critical insights into health symptoms like depression and anxiety. Our goal is to delve deeper into leveraging this data to furnish enhanced health insights as proactive measures against potential health ailments. Additionally, we aim to refine lifestyle recommendations for the user's calendar to foster better recuperation.