AWS Partner Network (APN) Blog
Building Next-Gen AI Education through GoML and Amazon Bedrock
By Rishabh Sood, Chief Executive Officer at GoML
By Prachi Doshi, Partner Solutions Architect at AWS
By Pyone Thant Win, Partner Solutions Architect at AWS
 |
 |
Generative AI (GenAI) is transforming education by providing students with personalized and adaptive learning experiences. AI-powered EdTech can personalize learning by analyzing individual styles and providing customized content like practice problems, concept summaries, and real-time feedback. GoML, an AWS Advanced Consulting Partner, is driving this transformation through its expertise in cloud computing, natural language processing (NLP), machine learning (ML), and large language model operations (LLMOps). GoML collaborates with EdTech firms to provide personalized content, real-time query answering, and dynamic assessments. GoML’s Amazon Bedrock-powered solution drives better learning outcomes for both students and educators globally. This blog explores how GoML’s partnership with AWS and a leading EdTech organization leverage GenAI to transform educational experiences.
Understanding the challenge
Students preparing for highly competitive exams like JEE Advanced, NEET, SAT, and Gaokao often face exceptionally complex questions. Manually solving these complex questions can be error-prone and time-consuming, leaving students frustrated and struggling to grasp intricate concepts. This inefficiency negatively affects a student’s preparation and performance in these competitive exams. Exam preparation requires a detailed and intuitive doubt solver to support students.
A top 10 Asian EdTech organization, known for its Educational Android application, is delivering exactly the solution for exam preparation for students in grades 6-12 and entrance exam candidates. The solution provides tailored practice modules, mock tests, previous year papers, and live doubt resolution sessions for students to practice. The organization leverages GenAI capabilities to offer personalized learning solutions to address diverse student needs from middle school to competitive exams. Funded by the Ministry of Electronics and Information Technology (MeitY) and the Government of India, the educational learning solution is at the forefront of India’s evolving education landscape, enhancing learning outcomes through innovative and technology-driven approaches.
The GoML solution
GoML collaborated with the EdTech organization to develop an AI solution for exam preparation. The solution leverages NLP to understand and interpret student queries by integrating Amazon Bedrock Knowledge Bases, advanced ML algorithms, and DSPy, an open-source Python framework. The Bedrock Knowledge Base of the solution consists primarily of question banks, sourced from textbooks and previous exam papers. It is a critical component of the solution, ensuring the breadth and depth of topics to cater to a wide variety of potential queries. Bedrock Knowledge Bases is a feature in Amazon Bedrock that implements Retrieval Augmented Generation (RAG) to achieve precise answers that enhance the learning experience for students.
A common technique, RAG takes data from different sources, transforms it into vector embeddings stored in a database, then uses these to enrich user queries with context. The process begins by vectorizing a submitted query, locating relevant document chunks via similarity search, and using this context to enrich the LLM prompt, generating the eventual response. The LLM leverages the vector database for informed, contextually relevant responses.
Amazon Bedrock is an ideal service for powering this educational offering because of its interoperability and access to a broad selection of foundation models (FMs). The seamless integration with various AWS services enhances workflow efficiency, while Bedrock Knowledge Bases’ RAG capabilities make sure that students receive contextually detailed responses. Its cost-effective pay-as-you-go model also helps allocate resources effectively and its responsible AI features, such as Amazon Bedrock Guardrails, enables GoML to implement protections against sensitive customer data.
Outcomes
GoML integrated advanced GenAI features into the existing solution. Using cloud technologies, the solution provides near real-time answers to student questions. Student satisfaction rose thanks to this support, enabling educators to prioritize engaging, advanced teaching methods. GoML incorporated reinforcement learning from human feedback (RLHF) technique to the solution which improves over time, adapting to students’ evolving needs.
Implementing GoML’s GenAI solutions on this education and learning solution has yielded the below outcomes.
- 95% accuracy in solving IIT, JEE, and Gaokao-Level exam questions
- Achieves high precision, offering students reliable and correct solutions
- Reduces uncertainty and frustration associated with complex problem-solving during exam preparation
- 80% reduction in effort during exam preparations
- Streamlines the learning process, allowing students to save time
- Improves the efficiency of study sessions
- Step-by-step solutions to reinforce difficult concepts
- Break down complex problems into manageable steps for better comprehension
- Expandable steps for more detailed doubt resolution and concept refresher.
Technical and architectural overview
Model choice – Meta’s advanced Llama 3.1 8B model, accessible through Amazon Bedrock, powers this solution. This decision-tree model, combined with Bedrock Knowledge Bases, simplifies challenging student questions from exams like IIT JEE, NEET, and Gaokao into clear, explainable answers for improved understanding.
Optimizer and evaluation – GoML used DSPy, an open-source programming framework, that simplifies the development and optimization of Large Language Model (LLM) applications through a higher-level declarative interface. It enables developers to compile natural language tasks into optimized LLM prompts and automatically tune their applications for better performance, essentially acting as a compiler between human-specified tasks and LLM calls.
The solution’s workflow begins with users accessing the environment through a secure web interface that’s protected by AWS Web Application Firewall (WAF), ensuring robust security against threats and unauthorized access. Amazon API Gateway effectively handles and directs these user requests to the correct backend services. Triggered by incoming requests, AWS Lambda functions process inference logic, providing real-time interaction within a serverless architecture that has shown a 30% increase in query response speed.
The system’s processing continues as Lambda functions communicate with FastAPI and integrate with LyzrSDK, using LLM with Bedrock Knowledge Bases that have enhanced inference accuracy by 25%. User documents are securely stored in Amazon Simple Storage Service (Amazon S3), while Amazon Elastic Compute Cloud (Amazon EC2) instances in a private subnet handle document processing and embedding generation using LLaMA2. The architecture leverages Amazon Elastic Container Registry (ECR) for containerized application storage, Amazon Relational Database Service (Amazon RDS) for PostgreSQL for structured data management (resulting in 40% increased data insights efficiency), and maintains security through AWS Secrets Manager and monitoring via Amazon CloudWatch. This comprehensive workflow ensures efficient processing, secure storage, and reliable performance monitoring throughout the entire system.
Figure 1 shows the GoML’s doubt resolution architecture on AWS with Amazon Bedrock.
Figure 1 – GoML’s doubt resolution solution architecture
Conclusion
In this post, we explored how GoML is revolutionizing educational technology through its innovative use of Amazon Bedrock and generative AI capabilities. The transformative potential of AWS in education is clear in GoML’s doubt resolution system—95% accurate on complex questions, 80% less student prep. Their system, built with Amazon Bedrock, Bedrock Knowledge Bases, and AWS services, highlights the potential of AI for creating effective, personalized exam prep for students.
Find out more about creating similar GenAI solutions using Amazon Bedrock or explore how GoML can revolutionize your educational technology by visiting GoML’s AWS partner page or contacting their team. Explore Amazon Bedrock’s features further via AWS documentation and begin creating your GenAI-driven educational tools now.
.
.
GoML – AWS Partner Spotlight
GoML is an AWS Advanced Consulting Partner with AWS Generative AI Services Competency. GoML provides LLM application solutions and services, boasting over 50 business-ready and 10 production-ready use cases across diverse sectors like finance, healthcare, education, and manufacturing. With their LLM Starter offer they are helping clients achieve a production-ready use case on the LLM of their choice in 8 weeks.