Inspiration
With the widespread adoption of artificial intelligence, many people have begun to create and implement their own machine learning models, but there is a serious issue: the datasets are often biased, imbalanced and poorly made. this leads to a final model that is inaccurate, which could lead to serious real world affects, which is why we want to make data balancing and bias detection easier to accomplish and implement.
What it does
our website takes in uploaded datasets and using a combination of techniques we provide a way to weight the files to ensure a more balanced dataset, along with an in depth statistical analysis, allowing the user to manually tweak the weights as needed.
How we built it
We used mongo DB for the database, html and JavaScript for the website front end, and python for the backend which used Gemini based agents that created the statistical analysis. We heavily relied on LLM coding assistance.
Challenges we ran into
we were able to quite easily create all the pieces of the project: Vultr for hosting, Mongo DB for the database, and all the scripts for the backend and frontend, but we struggled heavily trying to combine all of them into one working product. in the end we had to cut the Vultr hosting, and instead pivoted to localhost.
Accomplishments that we're proud of
The localhost works well and provides a good statistical analysis of the data provided. all the core features work as intended.
What we learned
We learned a lot about web hosting through vultr, utilizing NoSQL databases such as MongoDB through atlas, as well as integrating a front and back end for website development.
What's next for Skewd.AI
We are working on hosting using Vultr, as well as creating a way to augment and refractor the datasets that are uploaded to it to allow for more detailed balancing.
Log in or sign up for Devpost to join the conversation.