Sophic

Personalized AI content creation platform

Comment

Inspiration

Sophos is the greek word for wisdom. Sophic is intended to be the future of content consumption - the AI generated content is designed to be maximally interesting to the user but still have a lean towards stimulating their minds with knowledge, ideas, and creativity that they would otherwise not have been exposed to.

What it does

Displays a feed of personalized text and image generations based on the user's interests. The user data is sourced from the chat component of the app that generates images and text - Sophic records this user data and runs gpt3.5-powered analyses on it to inform the feed's AI content.

How I built it

I used React Native in the front end and Node.js in the back. I currently have implemented many gpt3.5 api calls, a few dall-e api calls, and AWS s3 storage and retrieval.

Challenges I ran into

Everything was a challenge to me given this was my first software experience but the challenges were fun to overcome - I haven't run into a problem I couldn't solve eventually.

One particular problem involved storing image generations. Dall-e api returns a URL with an hour expiration date making them unusable in the long term. I had to implement a system that breaks dall-e URLs into binary, reconstruct the images, and store the new image in AWS s3. I then had to retrieve the AWS s3 URL and save it in my app's asynchronous storage. This was a difficult task.

Accomplishments I'm proud of

This is my first software development project or any technical project. I am a philosophy major at UC Berkeley and have been teaching myself to code with ChatGPT over the past few months. Coming to this hackathon and beginning to build Sophic has been a monumental accomplishment.

What I learned

My highest priority should have been to focus on the quality of the feed's generated content. I was spending a lot of time with design and platform functionality that I lost the opportunity to prompt engineer even better AI content.

What's next for Sophic

Short term: implement semantic search on file names of all AWS s3 images to recycle the AI generated images and implement them into feed to avoid having to generate new images all of the time (user's wont know the difference and it cuts API costs drastically)

Longer term (next 3 months): implement login, include ads in the feed, and then offer a subscription to remove ads, to generate more images, and to replace GPT3.5 api with GPT4. I will launch by September this year.

Built With

Share this project:

Updates