Sunshine | Devpost

Website Layout

Inspiration

During BoilerMake VI 2018 Purdue, a 36-hour hackathon which welcomes all levels of experience and provides tools and resources, our team decided to focus on creating a convenient way to analyize the weather. After being able to talk to sponsors for BoilerMake and other students, we wanted to create a convenient way to analyze the weather. Often times, weather data can be overwhelming and hard to consider the factors of wind, precipitation, humidity, and temperature. Through Sunshine, we wanted to allow users to know what type of clothing to wear without having to complicate weather data.

What it does

A user-driven Alexa skill and web application that suggests clothing preference based off of weather and user's response to different weather conditions.

How we built it

These are the APIs and resources we utilized. The programming languages we used included Javascript, HTML, CSS, and Python.

MongoDB Stitch - Serverless platform with simple, secure access to data and services from the clients
TensorFlow.js - Library for training and deploying machine-learning models
Dark Sky API - Weather forecasts and data web API.
Alexa Skills - Amazon’s cloud-based voice service tool kit

Challenges we ran into

Developing something new, we ran into a lot of errors. For the Alexa, it was a challenge to find a way to request a POST from the endpoint. For MongoDB-Stitch, it was challenging to get accustomed to storing data in a new way. For TensorFlow.js, familiarizing ourselves with machine learning was new.

Accomplishments that we're proud of

Being able to create a product that can be published.

What we learned

Having never used MongoDB-Stitch, TensorFlow.js, and Alexa Skills, we learned how to utilize these tools and resources to our product.

What's next for Sunshine

We would love to see Sunshine grow outside of just Alexa and web application but also to mobile. Towards Alexa, we would want to create new commands to further allow convenience. Not only that, we can improve the modeling of the prediction service. To further scale the service, we would require a better way to manage the data base as well.