ARTIFICIAL INTELLIGENCE
Intelligent Search for Educational Content with AI
Modern artificial intelligence (AI) technologies are opening entirely new possibilities in the personalization of education. One particularly valuable application of AI is in automatically matching educational content to a learner’s individual needs — including through the analysis of visual inputs.
In this project, the team at ALTEN Polska developed an advanced educational content recommendation system based on the analysis of uploaded images — such as drawings, assignments, or diagrams. Leveraging the Gemini API, Text Embeddings technology, and Vertex AI Vector Search, the team created vector representations of content that could be effectively matched with educational resources. This resulted in over 90% recommendation accuracy and a significant reduction in the time required to locate relevant materials.
Challenge
Enable students to quickly find the most relevant educational content based on an uploaded image.
Solution
Create a hybrid content recommendation system that leverages AI and image-based vector analysis.
Benefits
- Over 90% recommendation accuracy
- Faster access to relevant materials
- Increased user engagement with the educational platform
- Scalable, fully cloud-based architecture
- Intuitive user experience and time savings
The role of AI in addressing the challenge
One of the major challenges involved transforming unstructured visual data (e.g., photos of math problems or handwritten sketches) into a system-readable format suitable for effective content matching. Another critical challenge was maintaining a high level of recommendation precision to ensure that users received truly helpful materials. This required the integration of image processing technologies, descriptor generation, and semantic search capabilities in the cloud. Eliminating the barrier of time-consuming manual resource searching significantly enhanced the overall user experience.
AI technologies and tools applied in the project
The project employed Gemini AI for generating image descriptors and vectorizing them. Vertex AI Vector Search was then used to implement a hybrid search mechanism that combines semantic matching with traditional filtering. The entire system was deployed on a Google Cloud-based architecture, ensuring scalability, high availability, and robust security. It was built from scratch and fully tailored to the client’s specific needs.
Tangible results – client and user benefits
Thanks to the implementation of the AI-based recommendation system, users of the educational platform gained instant access to the most relevant learning materials without the need for manual searching. The image-based search feature enables students to receive rapid learning support, boosting both engagement and learning efficiency. For the client, the system led to increased interaction with the platform and introduced a scalable, innovative technology solution capable of evolving alongside the platform and its users’ needs.
