Accedia Brings Responsible AI to the Classroom with AIcademy

Accedia has developed AIcademy, an educational platform that uses artificial intelligence to generate personalized learning materials for primary and secondary school students. The platform has been tested and demonstrated in a real educational environment and was developed with the support of the European Union under the program "Competitiveness and Innovation in Enterprises" 2021-2027.

AIcademy is built around three core components: AI-generated learning materials, tools for teachers, and an adaptive algorithm that builds an individual profile for each student based on behavioral data, test results, and teacher feedback. The platform generates lesson summaries, flashcards, multiple-choice questions, event timelines, and provides real-time support through an AI chat assistant restricted to working only within the context of approved learning content.

Model Selection and Architecture

Before development began, the Accedia team conducted a structured comparative evaluation of leading large language models, including various versions of GPT, Gemini, Claude Sonnet, and Claude Opus. The evaluation covered content quality and relevance, operational readiness and ease of integration, data protection compliance, and scalability. GPT was selected based on a documented scoring matrix.

The core architectural decision is a hybrid approach: the platform combines GPT with verification algorithms developed by Accedia, organized across three layers of control. The first is pedagogical adequacy: validating that generated materials are accurate, age-appropriate, and educationally useful. The second is grounding in approved sources: the model works only with pedagogically approved learning materials, not external sources or its own interpretations. The third is behavioral guardrails: a defined child-safe tone and clear rules for what the assistant can and cannot answer in a school context.

"The hardest part was not the model integration itself but defining its boundaries. We had to train the system to recognize when a question falls outside the lesson context and respond in a way that is appropriate for a child. Guardrails are not a feature you add at the end. They are an architectural decision that affects the entire product, " shares Bozhidar Goranchev, Head of LLM Integration.

Every generated material goes through teacher review and approval before it reaches the student. The AI accelerates preparation; the teacher controls the outcome.

Team and Process

Behind AIcademy stands a team of 16 specialists, including project leads, AI and LLM specialists, frontend and backend developers, a UX designer, a business analyst, and a QA engineer. Development follows an Agile methodology with two-week sprints and was preceded by in-depth user research with real teachers and students, whose conclusions are directly reflected in the product decisions.

About Accedia

Accedia is a European IT services company specializing in technology consulting and custom software development. Founded in 2012, the company works with clients across more than 20 countries on 5 continents.

This project is funded under Contract No. BG16RFPR001-1.001-0173-C01 under the "Competitiveness and Innovation in Enterprises" Program for 2021–2027, co-financed by the European Union.