Team Cloud 9ers is Mustafa Khan, Zuhaib Shakeel Khan, Phani Ram Teja Ravipati and Lakshmi Sai Kartheek Surabhi.
Abstract
The Smart AI Retrieval Assistant, nicknamed SARA, is an AI tool which would read your personal documents like a resume or class notes and answer questions from within those documents. To accomplish this, it uses 3 main components:
- A Vector Database
- An Embedding Function for the Vector Database
- A LLM for generative answering
For this project, the ChromaDB Vector Database was used, OpenAI’s embedding model ‘text-embedding-ada-002’ was used for embedding and for intelligent generative answering, we use OpenAI’s ‘gpt-3.5-turbo-16K’ model.
These types of chatbots are also referred to as Retrieval Augmented Generation (RAG) chatbot. They retrieve some data as context, in our case, the vector store, they augment it as a prompt by using the retrieved information as context, and then use the prompt to generate an answer.
We provided an early concept version of SARA to students from multiple majors studying in UW-Milwaukee, and we got a positive response and a need for such product in the interest of the students.
The most positive response came from Business Majors, who had a plethora of Power point files which contained source material, and they found SARA useful in answering their questions.