OpenRizz

Inspiration

At UTD a common problem students experience is loneliness, a simple fix for that is to socialize, but that's hard! So we created a platform using AI to help people overcome the anxiety of dating and reduce the rate of loneliness at UTD.

What it does

OpenRizz takes dating profiles of your match from any dating app as an input and enables you to talk to an AI tailored to that dating profile. OpenRizz also recommends the optimal outfits to wear for your date. It also takes in your information and dates to create a personalized set of possible date locations.

How we built it

We utilized NextJS for the front end and Flask for the back end. Our development process involved integrating AWS Textract and OpenAI API to extract keywords from profile images. This allowed us to create a feature where the user can practice conversation with another person before their date. Additionally, we employed Dalle to generate customizable clothes based on the extracted keywords for the wardrobe section of our application.

Challenges we ran into

The OpenAI API is a powerful tool for natural language processing, but it is not immune to issues and bugs. The main issue was the process of tokenizing keywords from images which appeared to be more challenging than anticipated. It required a significant amount of trial and error to fine-tune the extraction process and ensure high precision and recall. The variability in image quality and content added an additional layer of complexity to the tokenization process. Despite these challenges, we were able to successfully extract relevant keywords from images and utilize them to generate customized content for our application. While it required extra effort and time on our part, it ultimately allowed us to achieve the results we needed despite the challenges posed by the OpenAI API. In addition, a significant amount of time was dedicated to establishing a connection between the front-end and back-end using NextJS and Flask. As a result, we were able to create a seamless and integrated user experience for our application.

Accomplishments that we're proud of

Developed advanced prompt generation enabling the AI to behave like a real-life individual to make the mock dating experience as realistic as possible

What we learned

We recognized and learned the significance of preprocessing text in NLP. Therefore, we taught ourselves to implement a tokenization process to extract text from images, arrange the content by line, and then categorize it according to various attributes of the profile such as basic information, lifestyle, interests, favorite music, and more.

What's next for OpenRizz

The goal is to sell our application to dating apps such as Tinder, their vast database of information can help us make our AI even better. Not only that, Getting OpenRizz to become integrated into a dating app would enable its features to help reach more people and make the dating experience less anxiety-inducing for the user.