Aesop.ai

Avengers! When kids hear that they think FUN! Yet, when kids hear learning they Cry! AESOP uses generative AI to allow their favorite superheros into dynamic mentors to them about educational topics.

Comment

Inspiration

We are passionate about teaching and inspiring kids to succeed. After brainstorming, we discovered that we all had relatable frustrations of being unable to engage children to learn school topics from our tutoring experiences. All the kids would love to talk about the latest superhero movies and would rather listen and talk about them than to engage with us. Knowing that, we decided to harness the generative AI and let the heroes take over the teaching from us.

What it does

Utilizes multimodal AI agents to generate text and illustrations to create engaging and interactive choose-your-own educational adventures. The child picks his favorite character, his grade, and the subject he wants to learn so that they can build their own unique experience

How we built it

We utilized:

  • Groq's LiteLLM and LPU backend to generate text for our stories.
  • Reflex's pure-Python and batteries-included approach to Web apps to accelerate our front and backend development
  • Cartesia to build a clone superhero voice for superhero narrations.
  • Hyperbolic's Dall-E 3 to create imaginative images based on the text provided

Challenges we ran into

  • Implementing the vector database and content databases.
  • Getting around the Reflex's data modeling limitations
  • Implementing tracing with Arize Phoenix for child safety.

Accomplishments that we're proud of

We are proud of our end to end product. We are proud of how the we completed the process of how we completed the process of brainstorming to building a product in 36 hours. We can proudly say that we built a product that can potentially be used for an socially positive use case .

What we learned

We learned how there can be a lot of problems throughout the project process and we learned that persistencey is the key. We learned that communication is especially important in relaying and solving problems. As for the techinal side we learned a lot about how to use git along with learning how to string together different AI agents to become one successful product.

What's next for Aesop.ai

  • chromadb to create vector databases that store word embeddings and enable fast searches.
  • As Arize Phoenix's API had a bit of learning curve and child safety is an important issue worth deep discussions, Implementing tracing has become our next time priority
  • Accessibility features: Multilingual support, dyslexic and learning handicap features -Fetch.ai to connect to agent-verse

Built With

Share this project:

Updates