Inspiration
We got the inspiration by watching each other collapsing to finalize a hackathon idea at 5 a.m. in the morning. We've tried multiple other methods like hackathon idea generators online but didn't get much idea from those randomly arranged tech buzzwords. Then the divine inspiration came as we would like to develop a solution that solves our pain point on the spot.
What it does
Our product provides users, and hackers in all hackathons, an opportunity to generate a unique, challenging, potentially award-winning idea in filtered areas of interest. We made sure that the concept is highly potential and outstanding in the fierce competition of hackathons by using the NLP algorithm powered by co:here API to analyze all winner projects on the Devpost website.
How we built it
We powered up co:here API to read the input we scraped from past hackathon projects on Devpost. We fetch the information we need such as tools and descriptions. Then the output is shown in an interface designed by Figma and implemented by React.
Challenges we ran into
We have to limit the input and output size as the sum of those has a maximum limit. Hence, we have to find a way to balance the number of projects, and the amount of info we feed into the algorithm. Also, lots of bugs when trying to connect the frontend and backend.
Accomplishments that we're proud of
Everybody who heard the idea got a good laugh. Aside from the fun examples, we got a few project topics that sound interesting and doable in a hackathon environment based on feedback from multiple friends. We also managed to hook up the backend and frontend after endless debugging which allowed users to use the program without much learning effort.
What we learned
We got our hands wet in the NLP and ML area during this project. We understood more about the true potential of artificial intelligence after seeing this task that we struggle with so much can be solved this way. And we got some cool hackathon ideas in our banks!
What's next for New Project Title
We would like to add features to allow people multi-select more than one area of interest to serve better when the hackathon group doesn't have a clear mind about what topic to tackle. We also would like to train our own model using all the winner posts on the Devpost to get a better result.
Built With
- co:here
- figma
- natural-language-processing
- python
- react
Log in or sign up for Devpost to join the conversation.