Trendy Sidekick

Inspiration

People who have several massive, related data sets (for example, all the data about the servers you are using) often need to compare data sets when diagnosing problems. One data set may indicate that something is going wrong, and the reason may be found in another data set. Sometimes, the relationship between what's happening in each of the data sets is obvious. Usually it is not, especially when a subtle or obscure trend is visually displayed on a graph or in a table, where a very gradual change or the noisiness of the data makes it difficult to perceive what it happening. People who do not use sight to read the data find themselves scrolling through zillions of data points, trying to keep track of two or more different data sets in their head.

What it does

Trendy Sidekick is your sidekick in non-visual data analysis. Trendy Sidekick scans through the zillions of data points for you so that you can zero in on the relevant trends and behavior of your data. If one of your data sets indicates that something has gone wrong with your system, Trendy Sidekick will search for trends and correlations between all the related data sets in order to suggest manageable sections for you to investigate further.

How we built it

We discussed extensively with Chris and Santi their data analysis process and where exactly it is going poorly. From this we identified that if we could significantly reduce the amount of data they have to scan through when trying to understand correlations and other interactions between data sets, this would be a step in the right direction. This tool would do trend and correlation analysis using Python on data files such as .csv and .json. It would display output as text, using a python GUI or displayed in the program they are using to collect data (such as AWS or MATLAB). We chose wxPython for the GUI in the hopes that it will be friendly with screen readers and braille displays as advertised -- we tested it with a screen reader but this was not exhaustive testing.

Challenges we ran into

We lost more than half our team, which meant we needed to reevaluate what the remaining members wanted to get out of this experience and deliver at the end of the day.

Accomplishments that we're proud of

We talked to Chris and Santi a lot, and we learned a lot from them about what they do and what they wish were better. We went to a bunch of workshops and met new people!

Chris and Santi and everyone else: Awesome!

Thanks so much to everyone who talked to us. It was an amazing opportunity to meet people with great ideas and to take in their perspectives.

What's next for Trendy Sidekick

We think that Trendy Sidekick is a very feasible idea and a potentially very useful tool. There are many possibilities for making it a very powerful sidekick, such as implementing machine learning to get more nuanced analysis of trends and data behaviors. While we did not have the people power to get it up and running, we hope someone (maybe even one of us!) will find the ideas compelling and make it happen.