Feedom

System Diagram: Recommendation Algorithm
System Diagram: Data Processing

Inspiration

In today’s digital age, social media platforms like Instagram and TikTok have become integral to our daily lives. However, their algorithms often prioritize content that keeps us hooked, leading to a cycle of endless scrolling and diminished productivity. This addictive nature of social media can negatively impact mental health, reducing our ability to focus and engage meaningfully with the world around us. Introducing Feedom, a revolutionary social media platform that empowers users to take control of their content consumption. Inspired by the need to balance digital engagement with personal well-being, Feedom aims to create a healthier social media environment where users dictate their content journey.

What it does

Feedom allows users to experience the same engaging content from platforms like Instagram and TikTok but with a crucial difference: the algorithm is fully controlled by the user. Instead of letting the platform decide what to watch next, users can customize their feeds based on their moods and preferences. Whether they want to see uplifting content, educational videos, or specific interests, Feedom lets users tailor their content consumption to match their goals. This way, users can enjoy social media without falling into the trap of endless, mindless scrolling.

How we built it

System_Diagram_Data The first stage is how we gather data and processing it for better recommendation system. We have used web scrapers to gather posts / videos and reels from Instagram. The idea is to run the web scraper continuously and download more and more data for our platform. Once we have content downloaded on the server, our Machine Learning pipeline starts. We use Vision Transformers to extract out important information from the video provided and pass that data to a Large Language Model which will act as a tagging system. To make this task easy for us, we have used OpenAI's LLM. Our data processing system runs separately from our main module, the recommendation system.

System_Diagram_Recommendation Our recommendation system provides two level of personalization,

What user likes / prefers, this is handled by our platform. This recommendations are similar to what other platforms like Instagram, Tiktok, etc.
What user wants to see or what user should see.

Our app, at it heart, helps user detoxify the attachment from social media platforms. So, with the help of the second personalization algorithm, we can help user control their content.

Challenges we ran into

Scraping the user data from closed source applications like Instagram. Our initial plan was to scrape the user feed and refine their own content, but we need application's approval for that.
Figuring out how monetization will work in real-life as our whole aim of the project is to help user get away from social media platforms like ours.
Time crunch, implementing the whole app in the limited amount of time.