Duration: May 2024 - July 2024
Organization: Coriolis Technologies
During the summer of 2024, I had the opportunity to work with Coriolis Technologies on a project focused on automating the creation and grading of subjective questions and answers from PDF documents.
- Objective: Develop a system to generate diverse types of subjective questions from PDFs and automate the grading of answers.
- QA Generation:
- Created refined prompts for generating a diverse range of questions, enabling a deeper and non-monotonic understanding of the documents.
- Implemented RAG to generate contextually relvant answers
- Evaluation Tool:
- Developed a robust evaluation mechanism capable of assessing answers with a high level of precision.
- The evaluation system has been tested against human evaluations and LangSmith to ensure accuracy and reliability.
This project aimed to enhance the efficiency of creating and grading subjective questions/answers, providing a more comprehensive assessment of document content.