Project summaryPathlon
AI Peak Performance Coach
Client: Startup
Year: 2019
Skills: VUI Design, User Interviews, Technical Architecture, AWS, ML, Data Analytics
-
Create a way for athletes to get actionable feedback on their optimal time to work out on any given day.
-
A simplistic daily interaction designed with athletes’ need to balance peak performance and the psychological impact of time.
Pathlon imagines a future personal fitness coach, where athletes can quickly ask for tailored actionable tips. This AI coach, interfacing via Amazon’s Alexa, has been designed to remove hassle for athletes by recommending a personal optimal time of day to workout. It knows this by accessing your personal data from wearables and augments this with contextual and live environmental conditions such as humidity or temperature.
Interviews with athletes made us understand their need to know exactly what time of day they perform at their peak. The challenge was to uncover their peak athletic performance patterns. We wanted it to feel actionable and accurate.
Context
IoT has opened up new avenues for design and tech, allowing us to create novel services that can allow the user to access a vast amount of data in the form of actionable feedback. Personalised Healthcare, wearables, smart homes, and personalised assistants are all topics driving the conversation. We are now seeing the personal tracking industry explore new ways to enhance health and well-being.
This project aims to use your wearable data to inform the best time of day to train, by tracking; workouts, sleep, amount of calories burnt, and weight. Where we married this with training goals, personal calendars, and environmental data to determine the best time of day for the athlete’s training sessions.
To achieve our prototyping goals we knew we needed to deliver an accurate prediction model, that could solve the problem of finding your optimal time of day to workout. And more importantly, to easily interface with it. To create our prediction model we had to put in the prep work. This consisted of sourcing, logging, cleaning and analysing user wearable data. In parallel, to consider the future scaling needs of the startup, a serverless architecture was designed and set up using AWS to host and run the live prototype.
Over the course of a month, we logged and stored wearable data from every single athlete workout session, alongside hourly environmental data from local weather API. This data was then used to compile our initial test prediction model.
