The problem Acharya: Personal AI Tutor solves

Education in our country is unfair at large

• Undemocratic access to learning and personal/group tutoring
• Economic challenges faced by the underprivileged with regards to competitive exams and board/national exams
• Lack of mentorship and knowledge about the same
• Lack of personal attention in the coaching and tutoring businesses

A personalised AI tutor can solve these issues at large, and provide knowledge and guidance to the underprivileged. Being from lower middle class families, we have often felt these problems first hand. A personalised, democratic, knowledgeable AI assistant for tutoring can help fix this education gap between the economic classes.

This is our aim with Acharya.

Challenges we ran into

GPT-4 can be highly unpredictable at higher temperatures. Additionally it has a lot of latency as well, and is quite expensive compared to older models. For text to speech and speech to text, we also faced similar latency from Eleven Labs.

Adjusting the prompt to suit learning and teaching commands was the most difficult part, and we did a lot of prompt engineering and testing for the same. For latency, we have tried to optimise the distance between servers geographically, and use Firebase's Realtime Database to maintain connections to the server.