Kipitup: Utilizing biosensory data for nuanced health and fitness recommendations

About Kipitup

Kipitup is an Apple Watch app that provides fitness advice for any athlete, no matter where they are on their journey. It delivers fitness advice that incorporates both the wearer's biosensory feedback and industry leading best practices, ensuring that advice is always specific to what would be best for the wearer. Kipitup is able to provide specific steps for improvement and feedback for 'cardio training' and ‘marathon preparation’.

User interactivity

Users will interact with Kipitup using by asking natural-language questions to their Apple Watch. After Kipitup takes into account the user's biofeedback and trends, it will reply with customized user-specific results. Those results are catalouged in the app so users can see historical advice and instructional plans.

How Kipitup works

Users vocally ask a fitness-related question using the Kipitup app in their Apple Watch. Kipitup converts their question using speech-to-text. The user’s question is directed to a large language model (LLM) we have trained to provide industry-leading fitness advice. In parallel, a biosensory ‘snapshot’ of the user is taken using the Apple Watch. Kipitup sends the biodata to the LLM (stored in a AWS Lambda instance) which processes it and converts it into a machine-readable string. The processed biosensory data is also directed to the LLM to provide additional context about the user. Finally, Kipitup generates fitness advice based on existing best practices which are tailored to the user’s biodata, ensuring that recommendations are always unique and customized to the user. That generated fitness advice is then sent back to the user via email.

Download Kipitup

Kipitup is currently available as a locally downloaded file. Instructions for how to load Kipitup to your Apple Watch are in the 'Final Paper' document. The download links below access a Google Drive in which these documents can be downloaded directly.

Kipitup App (.zip):          [Download]

Final Paper (.docx):        [Download]