Case Study – Generative AI

Renowned EdTech company – Video Summarization using Amazon Bedrock

About the customer

It is premiere training institute pioneering in individual segments such as IT learning solutions, banking & finance and English Communication. With updated training methodologies and certified trainers, Aptech Learning offers its students, a competitive edge in their careers enabling them to become dynamic job-ready professionals with its USPs such as employability enhancement program, holistic training infrastructure, placement assistance, live projects, new age course curriculum, certified faculties and Onlinevarsity-an exclusive e-learning platform

Challenges

Use Case: GENERATIVE AI – using Amazon Bedrock

The institute providing training in various fields such as IT, Banking and Finance, Art, and English Communication, faced significant challenges in creating training materials for their video learning courses. Each course comprises approximately 20 modules, with each module containing 10 to 12 videos. Faculty members were manually preparing handout documents for each video, which presented several challenges:

  1. Time-Consuming Process: Faculty members were spending excessive time creating handout documents, diverting their focus away from their core teaching responsibilities.
  2. Inconsistent Quality: There was a lack of uniformity in the format and quality of learning materials across different videos and modules, affecting the overall learning experience.
  3. Scalability Issues: The manual creation process struggled to keep up with the growing volume of video content, making it difficult to scale efficiently.
  4. High Costs: The use of third-party transcription tools was proving to be expensive, further adding to the resource strain.
thin

Solution

Operisoft proposed a comprehensive solution to address the customer’s challenges by developing an automated video summarization workflow. This solution leverages multiple AWS services, including API Gateway, Lambda, Transcribe, and Bedrock, in conjunction with the FPDF library.

The process begins with user upload video/PDF/word files to an S3 bucket, which triggers the automated workflow. If it is video file then AWS Transcribe converts these video files into scripts. If it is word/PDF file then using respective python library, extracted the text from it. These scripts/extracted texts are subsequently processed by AWS Bedrock, which uses Jurassic Ultra, a state-of-the-art large language model, to generate concise learning objectives, synopses, and key points. The prompts for the large language model have been meticulously crafted to ensure consistent output.

The final generated content is stored in PDF format using the Python FPDF library, ensuring that all materials are neatly organized and easily accessible. This streamlined approach significantly reduces the time and effort required to create training materials, enhances consistency across documents, and lowers operational costs.

AWS Services used: API Gateway, Lambda, AWS Transcribe, and Bedrock, S3 storage

thin

Solution Diagram

Outcome

The provided solution helped the customer with a more efficient, trustworthy, and secure approach to sharing the files between team members authenticated by Active Directory, using Native features of Active Directory for authentication and permission management. This not only gave the customer confidence in the solution but also peace of mind for his critical design data and BIM collaboration.

Additionally, as he knows that the FSx service is scalable, reliable, he can scale the storage space as well as throughput as and when the need arises.

  1. Time-Saving: The automation of training material creation saved significant manual effort, increasing efficiency by 70%.
  2. Consistency: Ensured uniform quality and format for all training documents, enhancing the learning experience.
  3. Scalability: The solution handles large volumes of content without extra effort, accommodating Aptech’s growing needs.
  4. Cost-Effective: Reduced operational costs by 40% through the efficient use of AWS services.
  5. Flexibility: The solution adapts easily to generate various types of summaries and insights, meeting diverse training requirements.
  6. Transcription Accuracy: AWS Transcribe met the needs for precise transcription of different pronunciations and technical languages near to 95%.
  7. Cost Efficiency: Leveraging AWS’s serverless architecture and pay-per-use services further reduced costs.
thin