What was the goal?

An educational institution faced the daunting task ofmanually creating thousands of doctoral-level questions derived fromproprietary study materials. These materials consisted of two main components:

1. Static   Content (25%): A portion of the study material remained unchanged from previous years, requiring only minor revisions to questions.
2. Current Affairs (75%): The bulk of the material was subject to rapid changes based on contemporary events, necessitating the constant generation of new questions

The primary challenges in question generation were:

1. Time Constraints: Manual authoring of questions was time-consuming.
2. Cost:  The labor-intensive nature of question authoring made it an expensive endeavor.

To address these challenges, the institution required asolution capable of generating doctoral-level questions, complete with answersand explanations, in a specific format, drawing from a variety of studymaterials.

‍

Solution:

Our solution leveraged state-of-the-art technologies,combining GPT-4 and the LLaMA2 architecture to create an automated question,answer, and explanation generator. We developed a user-friendly interface tofacilitate ease of use.

‍

Results:

The results were remarkable:

• Accuracy:  Our models achieved over 95% accuracy in question generation.
• Efficiency: We achieved a 50x improvement in the time required to author questions.
• Cost Reduction: Our solution delivered a 10x reduction in the cost     associated with question generation.

Moreover, our implementation offered the added benefit ofenabling the institution to expand its reach from a local to a nationwideonline presence, thereby enhancing its accessibility and impact.

‍

Summing up

By harnessing the power of AI, we not only addressed theinstitution's immediate needs but also provided a transformative solution thatsignificantly improved efficiency, reduced costs, and expanded opportunitiesfor growth and outreach.

‍