Inspiration
Instameet was inspired by the idea of bringing people together through a platform that enables seamless communication, interaction, and content sharing, making it easy to connect with like-minded individuals.
What it does
Instameet is a social platform where users can create profiles, share posts about activities they want to engage in for a specified duration, and connect with others with similar interests. Activities can span a range of categories such as: Food/Drinks Walk Games Chat Study Travel Movie Party/Dance/Music Users can explore these posts and join activities that interest them, making it a great way to meet new people and have shared experiences.
How we built it
Our application uses a custom dataset generated with a Large Language Model (LLM), which includes images, descriptions, and hashtags. Multiple user profiles were created through random sampling to simulate diverse user behaviors. Embedded vectors are formed from the content (images, text, hashtags) and matched against user-specific vectors (based on interests and history) using a multimodal neural network. The prioritized content is then ranked using a decision tree to personalize the feed. The dataset comprises 1,000 samples.
Challenges we ran into
Creating a dataset using LLMs was complex and time-consuming. Ensuring diversity in images, descriptions, and hashtags required extensive fine-tuning of prompts and careful sampling. Balancing user preferences and historical data in the multimodal neural network required extensive experimentation and optimization. How we had to create a database and connect it to the website.
Accomplishments that we're proud of
We utilized LLMs to generate a diverse and comprehensive dataset of images, descriptions, and hashtags, simulating realistic user behavior. Developed a robust neural network for prioritizing posts based on user profiles, ensuring a highly personalized feed experience. Demonstrated the power of combining LLMs and neural networks to create a novel and engaging social platform experience.
What we learned
We learned how to generate diverse datasets using LLMs and integrate multimodal data (text, images, hashtags) into unified representations. We gained experience building personalized recommendation systems that adapt to user preferences and effectively rank content. We also tackled challenges like overfitting with small datasets and optimizing model performance. This project enhanced our understanding of combining AI techniques to solve real-world problems and deliver personalized user experiences.
What's next for Instameet
As a user, you can look forward to a more personalized experience with even better activity recommendations tailored to your interests. We’ll be enhancing our algorithm to suggest activities that align more closely with your preferences, making it easier to find like-minded people and shared experiences. Staying in touch with the people the met for those activities by using connect feature and also introducing for big groups.
Log in or sign up for Devpost to join the conversation.