Inspiration
The majority of our team members come from some form of data science/machine learning background. The one thing we always dread when starting a new kaggle competition or side project is scouring data and comparing metrics to find the perfect subset of features.
Our eyes bleed, our fingers crack, our souls slowly shrivel to wisps as we plug endlessly away at a keyboard like a monkey writing unexciting lines of the same library calls. There has to be a better way.
What it does
PandaFlex is a SaaS that provides a GUI abstraction to traditional data cleaning tasks. Through the interface, users can view correlation coefficients of attributes, various metadata, and mutate the data with just one or two mouse clicks. Once you're happy with the subset you've crafted, you can download the new data with the click of a button and jump right into training your models to do whatever you might imagine.
How we built it
The backend is a Flask REST API constructed through traditional Pandas approaches to data cleaning. We made sure to include correlation coefficient calculations, normalization ability, and even outlier/null value filtering. The works.
All this is presented in an adorable react frontend that will make sure there's no downtime as you cull the irrelevant aspects of your data.
Challenges we ran into
Design was a heavy challenge for this project. Understanding how to serve up a Pandas dataframe through a stateless web server was interesting to mull over and deciding on a UI that would not be overbearing was essential.
Accomplishments that we're proud of
Finishing our MVP within 12 hours!
What we learned
You can code really fast when you're sleep deprived.
What's next for PandaFlex
There's immense levels of complexity in data cleaning and it is disingenuous to think that what we've achieved in the past 12 hours can cover the task entirely. We want to improve on our UI and add additional functionality from data visualization to bootstrapped classification and everything in between.
Log in or sign up for Devpost to join the conversation.