The problem Superbook AI solves

Superbook uses Generative AI and technology to write a personalized book according to user's interest. Superbook provides a well managed productivity book reading interface that is powered by AI and technology.
Problems its solving:

1. A teacher or a presenter can use superbook to generate notes before presenting it to a class.
2. These notes can be a boiler plate for any Subject and user can expand the notes according to topics they require.
3. Students can use this to generate books even 1 day before the exam because of GPT-3.5 the book can be summarized depending on the needs of the readers and there time.
4. If reader don't understand something they can highlight and click on the "Explain Me" button to see explanation which will be generated in context to the Book.
5. Superbook has powerful backend search algorithm that searching the complete book. This algorithm can be upgraded to even search topics and user saved notes based on Natural Language so that user will have an experience similar to talking to a book.
6. Book can be written based on users prerequisite knowledge in the subject.
7. It contains AI generated Images to make the book interactive.

Challenges we ran into

1. Initially, our approach involved making repeated calls to GPT APIs to generate content, followed by storing it in the database. This process took more than 2 minutes to complete, resulting in a poor user experience. Understandably, users are unlikely to wait for such extended periods, so we recognized the need for a solution.
   One of the innovative solutions we implemented from the frontend was to initiate asynchronous book generation and employ long polling until the book was ready. Once the book was generated, the page would automatically load the content, significantly improving the user experience.
   To further enhance backend API performance, we implemented parallel processing of APIs wherever feasible, resulting in a substantial reduction in API response time to just 30 seconds.
2. When dealing with Generative AI the AI can have high hallucinations if the prompt is not enginnered properly. We improved the prompts and had to create validator and parsers to parse data from Generative AI because AI sometimes gives data in random format.
3. As a team, following a particular tech stack was a bit challenging, as each member had an experience on different tech stack.