Gen AI-based learning management system development for a Canadian Edtech company

Team Daffodil developed a smart & sophisticated text parser that automated the entire process of syllabus extraction. The system would pull data from the pre-uploaded syllabus files in a matter of seconds, which would earlier take hours to do so.

The parser scanned the syllabus files and showcased the required details such as course name, subject, objectives, etc.

Another advantage of the solution was that traditional keyword-based search methods often missed out on semantically similar content, however, our developed solution ensured that content with similar meaning could also be found, even if the keywords were not the same. Let’s say an educator searches for “Web Development”, the solution would also showcase syllabi containing words like “Site Building” or “Website Creation”.

This would ensure that no related topic is missed and help educators to build more reliable teaching material.

Our client had a huge collection of course material, stored in various formats (such as PDF, DOCX, etc.) and locations (such as local drives, emails, etc.). Extracting the right information from those files was a challenge for the client.

Our developed solution automatically organized and loaded course material to cloud storage where it could be easily accessed and used by AI tools. This helped the AI understand and interact with the course materials in real time, making the whole process faster and more accurate.

The system also kept the files organized, even if they were in several layers of folders, making it easier to find specific content.

To implement a powerful search and recommendation system, our team utilized advanced prompt engineering techniques for Retrieval Augment Generation (RAG).

When a user makes a query, the system searches its internal database as well as external sources such as online knowledge bases and databases, to retrieve the most relevant and up-to-date information. Next, it generates a personalized, context-aware answer for the same utilizing the retrieved and available information.

This approach was adopted to enhance course content discovery and relevance, which was a significant concern for our client, as they needed to ensure that their educational resources were high-quality, up-to-date, and context-aware.

How this helped:

Rank-wise Textbook Recommendations: To ensure that educators have the most popular book suggestions handy, our team utilized Vertex AI Search to suggest the top 3 textbooks that align with the syllabus.

Automated Course Outline Generation: The solution generated course outlines from syllabus data helping educators with a quick peek into what the course is all about.

Content Generation: The solution’s capability to extract similar content from the available files enriched the search experience for educators. It offered them with comprehensive content on similar topics. This way they won’t have to scrounge through multiple books and spend a lot of time in curriculum development.

This RAG-based approach not only improved the quality of the generated content but also enhanced the overall user experience for educators.

Additionally, Team Daffodil evaluated the performance of the RAG approach using the recommended methodologies of Google Cloud Platform (GCP) and tracked key metrics and performance indicators.

Our team worked closely with the client to define UAT criteria & test the solution. This collaborative approach helped to improve the solution’s performance and ensured the system’s reliability. We also conducted comprehensive training sessions covering system functionalities, best practices, and troubleshooting techniques.