Problem

Professional development can take teachers deep into new training material, but the real work often starts after the course content is delivered. Educators still need to find the right resources, prepare classroom materials, translate theory into practical lesson plans, and answer subject-specific questions with confidence.

Learning Assistant Platform set out to solve that problem with Unfold, an AI-powered learning platform for teacher professional development. At the center of the product is an AI voice agent for LMS experience: teachers can use intelligent assistants to navigate learning materials, create lesson plans, and strengthen subject-specific knowledge through personalized support.

The client brought Itera Research in at an early stage to validate the concept, build a proof of concept, and design the technical foundation for multiple AI agents working inside one educational platform.

The main challenge was clear: reduce the time teachers spend preparing lesson plans while giving them reliable support for platform navigation and subject-specific questions in one place.

Building an AI voice agent for LMS teacher development

Unfold was not meant to be a standard LMS with a chatbot attached. The client needed a platform where AI could actively support the teacher’s learning process without making the experience feel complicated or fragmented.

Teachers needed to be able to:

• find the right training materials faster;
• ask questions about courses, certification requirements, and platform functionality;
• generate lesson plans using a proven instructional model;
• get mathematics support through a dedicated knowledge base;
• interact with the platform through both voice and text.

Because the concept involved several complex AI interactions, Itera Research started with a proof of concept. This allowed the team to validate technical feasibility, test the user experience, and reduce risk before full-scale implementation.

To bring the platform to life, Itera Research followed a structured AI implementation workflow.

1. Designing a multi-agent learning ecosystem

Rather than rely on a single general-purpose chatbot, the platform was designed around three specialized AI agents, each responsible for a specific educational task.

One agent helps teachers navigate the platform and access relevant learning materials. Another supports lesson planning. A third helps educators better understand and explain mathematical concepts.

This approach makes the assistant experience more relevant. Instead of forcing every question through one broad AI layer, the system routes teachers toward the agent and knowledge base best suited to the task.

It also gives the platform room to grow. New learning domains, agents, and knowledge bases can be added later without rebuilding the entire product.

2. Building a lesson planning workflow around the 5E model

Lesson preparation was one of the biggest time-saving opportunities for the platform.

Itera Research built an AI Lesson Planning Agent around the 5E Instructional Model: Engage, Explore, Explain, Elaborate, and Evaluate. The agent guides teachers through objectives, activities, materials, and assessment ideas so they can move from training content to classroom-ready plans faster.

The goal was not to replace the teacher’s judgment. It was to give educators a structured starting point that follows pedagogical best practices and reduces repetitive preparation work.

3. Creating subject-specific support through dedicated knowledge bases

Teachers also needed support beyond general platform navigation.

The AI Mathematics Support Agent uses a dedicated educational knowledge base to provide contextual explanations and alternative ways to approach mathematical concepts. This gives educators practical help when preparing lessons or explaining subject matter in the classroom.

By using Retrieval-Augmented Generation (RAG) and dedicated knowledge repositories, the platform can deliver more accurate and context-aware responses than a generic assistant would provide.

4. Choosing voice technology that could scale

Voice was one of the key product goals from the beginning because the client wanted the LMS to feel more like a guided learning assistant than a static course library.

The team evaluated several voice technologies, including ElevenLabs, which is known for high-quality speech synthesis. AWS voice services were selected because they offered the right balance of scalability, security, and integration infrastructure for the platform’s long-term needs.

The platform now supports both voice and text interactions, allowing teachers to use whichever mode fits the moment. They can type when they need precision or use the AI voice agent for LMS navigation, course questions, or lesson preparation when they are moving through materials, thinking aloud, or working more conversationally.

5. Optimizing performance for real conversations

Early testing revealed response times of up to 10 seconds. That delay created friction, especially for voice interactions where timing affects whether the conversation feels natural.

Itera Research optimized agent workflows, retrieval processes, and infrastructure components to improve the experience. The team also refined model behavior to support more natural speech while keeping the architecture cost-effective for future growth.

The AI capabilities are orchestrated through Amazon Bedrock, with multiple specialized agents connected to dedicated educational knowledge bases. This AWS-powered architecture gives the platform a scalable foundation for pilot deployments and future expansion.

The resulting platform successfully completed preparation for pilot deployments in Saudi Arabia and the United Arab Emirates. During that phase, the team continued refining agent behavior, voice quality, and response times based on testing and user experience goals.

Early user feedback confirmed strong interest in AI-assisted professional development and classroom planning workflows.

Key Features

Multi-Agent AI Architecture: Three specialized AI agents support platform navigation, lesson planning, and mathematics education.

AI Learning Assistant for Platform Navigation: Teachers can ask questions about courses, certification requirements, learning materials, and platform functionality through voice or text.

AI-Powered Lesson Planning: The lesson planning agent helps educators create structured lesson plans using the 5E Instructional Model.

AI Mathematics Support Agent: A dedicated mathematics assistant provides contextual explanations and alternative approaches for subject-specific teaching support.

AI Voice Agent for LMS: Teachers can use voice interactions for platform navigation, course questions, and lesson preparation inside the LMS experience.

Voice and Text Interactions: Teachers can communicate with the AI assistants in the format that feels most natural for the task.

Educational Knowledge Bases: Dedicated knowledge repositories support more accurate, context-aware answers across learning domains.

AWS Bedrock Integration: Amazon Bedrock orchestrates the platform’s AI capabilities and supports the deployment of multiple specialized agents.

Modular Agent Design: The architecture is built to support future agents, learning domains, and knowledge bases without major platform changes.

Tools & Technologies

• React
• Python
• FastAPI
• PostgreSQL
• Amazon Bedrock
• AWS Knowledge Bases
• AWS Voice Services
• AI Agents
• Voice AI
• AI Voice Agent for LMS
• Large Language Models (LLMs)
• Retrieval-Augmented Generation (RAG)

Thinking of Building an AI Learning Product That Feels Practical?

Whether you’re building for education, training, or professional development, your platform can do more than deliver content. It can guide users, answer questions, and support real work in the flow of learning.

Get started today with a consultation.