Challenge
A customer approached us with a request to develop a platform for conducting training and testing knowledge on any topic independently.
The entire process – interactive discussion, and evaluation – needs to be powered by Artificial Intelligence to ensure maximum automation and independence.
The key challenge was to create an AI-driven solution where users could test their understanding through a live dialogue with AI instead of interacting with a real person. The platform needed to allow users to upload a document on a chosen topic, and conduct an interactive Q&A session on the topic from the document. Throughout the conversation, we should assess the user’s responses, providing a real-time evaluation of their knowledge.
Solution
Based on the client’s request, we built a platform where it is possible to conduct training on any topic using artificial intelligence. To bring this vision to life, we choose OpenAI’s GPT generative AI and the prompt tuning approach to improve the interaction with it and achieve a better result – a more natural dialog with the user.
We have built the following process of working on the platform:
- Document upload & Question generation
The user uploads a document on their chosen topic. AI then analyzes the content and generates relevant questions to test their knowledge. - Customizable training scenarios
Users can prepare a scenario with the questions they would like to test their knowledge, add context for the AI, or edit the questions as they wish. - Interactive Q&A session
The core of the platform is a Q&A session with AI in text or audio format. For the audio dialog, we used Azure AI Speech services to convert text to speech and speech to text. The user can answer his questions on the go in a language convenient for him and receive a score for each question in real time. - Performance review & Analysis
At the end of the dialog, users can view their results again for analysis, or browse the archive of previous results for comparison.
AI Optimization & System architecture
To ensure maximum efficiency, we implemented three specialized AI interaction classes:
- Document analysis & Question generation – Extracts key information and generates relevant questions.
- Live AI-driven dialog – Engages users in dynamic Q&A interactions.
- Answer evaluation – Assesses responses and provides real-time scoring.
For seamless integration, we built dedicated APIs for these AI functions, ensuring smooth, automated knowledge assessment and training.
Our solution transforms the way users prepare for discussions, negotiations, and knowledge testing – offering an intuitive, fully autonomous learning experience powered by AI.
Software architecture
The system has a single-page web application integrated with a monolithic backend application and supplemented by specialized modules for dialog management and AI data processing.
The system has the following 3-party integrations:
- Generative AI model (OpenAI GPT)
- Speech Service (Azure AI Speech)
The full project architecture and service interaction are shown below.
Key features
- AI-powered question generation to test knowledge based on user-uploaded documents.
- Creation of scenarios with predefined questions and input data to ensure AI interactions align with the user’s desired tone.
- Conducting AI-driven dialogues in text or audio format based on prepared scenarios.
- Support for dialogue in two languages: German and English.
- Dialogue evaluation with individual scores for each response and an overall knowledge assessment.
- User management system to allocate access across different companies or departments.
- Adaptive interface for seamless use on any device (desktop, tablet, or phone).
Results
As a result, we have implemented a platform that allows the client to conduct knowledge training on any topic of their choice. Thus, the client can prepare for any negotiations independently, without involving a second person, and get an assessment of their knowledge for further improvement. The solution significantly streamlines the learning process and enhances preparation efficiency.