Artificial Intelligence in Education: Digital Profiles, Avatars and Personal Learning Paths

In response, technologies are emerging that seemed like science fiction just five years ago: a digital profile that collects dozens of student parameters in real-time, and a teacher avatar capable of delivering a lecture in 70 languages, indistinguishablely mimicking the facial expressions and voice of a live mentor.

Russian and foreign projects are already transitioning from experiments to mass implementation, and it is these two areas - profiling participants in the educational process and creating "digital twins" - that are shaping the new face of schools and universities.

Let's take a closer look at what lies behind these technologies, what experience market leaders have accumulated, and what challenges we face on the path to AI-driven education.

Digital Student Profile: More Than Just Grades

A digital student profile today is a multidimensional model that combines cognitive, behavioral, emotional, and contextual characteristics.

The traditional portfolio with a list of grades and Olympiad results is becoming a thing of the past. Modern systems based on NLP and machine learning analyze written assignments, platform activity, task completion rate, and even emotional engagement to dynamically adapt the educational pathway. This approach allows for the formation of not just "academic performance," but an individual student portrait that helps teachers make decisions and students see their strengths and weaknesses.

Russian Experience: MES and "Student Portfolio"

One of the largest Russian projects that has already implemented AI profiling is the "Moscow Electronic School" (MES).

Since 2025, the system has included an AI service for mathematics, which analyzes a student's digital footprint, identifies knowledge gaps, and automatically suggests exercises to address them.

According to the developers, the service covers 1.2 million students and is capable of building an individual learning trajectory, taking into account topics that caused the most difficulties.

A digital profile is formed not only based on academic performance. MESH takes into account participation in olympiads, projects, and events, and in 2025 a career guidance module was added to the platform, helping schoolchildren decide on their future profession based on their interests and achievements.

International models: from NLP to conversational companions

On the international stage, the demand for personalization also drives the emergence of hybrid profile models.

One example is the integration of the IMS-LIP (IMS Learner Information Package) standard with generative AI. The model additionally takes into account psychological characteristics, for example, personality type according to the MBTI (Myers–Briggs Type Indicator) and learning style according to the Felder-Silverman model, while ChatGPT or Gemini help interpret this data and generate recommendations for teachers.

In a study by ECNUClaw (China), a five-dimensional student profile has been proposed, including cognitive, behavioral, emotional, metacognitive, and contextual dimensions. It is used in a conversational AI companion that supports schoolchildren throughout the learning process.

The Bakame AI project, which has won an international award, has even deployed an internet-free system to build profiles and adapt content for students in remote regions. Bakame AI is an innovative voice AI tutor that allows users to learn English via regular phone calls. The project was developed to make quality education accessible to 2.9 billion people around the world who do not have access to smartphones or high-speed internet.

Benefits and risks of personalized education

Just like students, teachers in the digital ecosystem are increasingly represented by their own profile, which records not only formal credentials, but also level of proficiency with AI tools, teaching preferences, and ethical principles.

Creating such a profile is now viewed as a key element of professional development, since the teacher is the one who brings AI technologies into the classroom.

Russian realities: “Teacher Assistant” and the legal framework

A landmark event for domestic education was the launch of the “Teacher Assistant” service, registered in the domestic software registry (entry No. 27417). The service provides feedback based on complex metrics, analyzes teacher performance, and generates recommendations to improve lesson effectiveness.

At the same time, the legal and ethical aspects of building a teacher's digital profile are being discussed in the professional community. Roundtables at Tyumen State University and publications in specialized journals emphasize that the profile should not become a tool for total control: a balance between analytics and professional autonomy is needed.

Today in Russia, the digital teacher profile is most often used in the context of professional development and AI competence assessment. There are already examples of its integration with LMS to select individual learning programs.

Foreign models: competency profiles and AI literacy

International practice is moving towards standardization.

In Spain, a study of primary and secondary school teacher profiles identified three clusters: teachers with high technological knowledge, high levels of pedagogical application, and strong ethical values. The ideal profile combines all three components.

The DAIC (Digital and AI Competences) professional standard for K-12 educators, proposed in Frontiers in Education, frames teacher AI literacy as a combination of technological cognition, pedagogical application, and ethical guidance.

UNESCO IITE is actively promoting training. In 2025, two Russian-language online courses were launched — "GenAI as a Teaching Tool" and "Prompt Engineering for Educators" — which have already been taken by thousands of learners. Today, in many countries, up to 40% of teachers admit they do not have sufficient skills to work with AI.

Teacher avatars: when a digital twin delivers a lecture

The most prominent trend of the past two years is the emergence of teacher avatars. Essentially, these are synchronous or asynchronous digital twins that copy not only appearance and voice, but also lecture delivery style, gestures, and emotional intonation.

Unlike static video recordings, avatars can interact with students in real time, answer questions, and adapt their delivery of material.

The time savings for teachers are enormous. By recording a course once or several basic dialogue branches, a professor gets a scalable assistant available 24/7.

Russian and foreign use cases for AI teacher avatars

Experiments with digital twins are actively underway at domestic universities. At the Graduate School of Management of St. Petersburg State University (GSOM SPbU), under the leadership of Olga Alkanova, educational videos with digital avatars were created that delivered lectures to students of business programs. Research showed that listener engagement does not decrease provided that "academic identity" is preserved — students want to see traits of a real instructor in the avatar.

RUDN University has developed a technology that allows creating a full-fledged digital twin based on a 30-minute video recording and text content. The system clones voice, facial expressions and communication style, after which the avatar is able to generate lectures on any topics.

Another interesting example is Moscow State University, where a scientist avatar trained on a professor's publications is able to hold discussions and answer students' questions as part of a scientific seminar.

International experience is no less impressive. At Imperial College Business School, the introduction of faculty avatars was driven by the desire to preserve the school's academic identity when scaling courses. Avatars created based on real professors not only delivered lectures, but also generated feedback in the LMS, and student surveys showed a high level of trust in such twins.

At Georgia Tech, professor David Joyner, one of the pioneers of online education, created his own digital twin that is able to update lecture content in real time, adapting to current discussions on course forums.

Berlitz language school, with a history of more than 146 years, has gone even further. Together with the Tavus platform, they developed empathetic AI instructors that do not just reproduce material, but also pick up on the student's emotional state, adjusting the pace and communication style. And at RISEBA University in Latvia, they launched AI Professor — a digital twin available for individual consultations via chat and video calls.

Technology stack

The key question is what all these avatars are built on.

Market leaders — HeyGen, Synthesia, ElevenLabs, Tavus — offer comprehensive platforms where you can generate an avatar in a few clicks, sync lip movements with an uploaded script, and get videos in dozens of languages.

However, the cost of such solutions for educational organizations with a large number of instructors remains high. Therefore, open-source projects are developing in parallel: for example, the "Talking Slide Avatars" framework, which uses a single portrait and script to sync speech and slide animation; or ALIVE — an interactive lecture engine where an avatar can answer student questions by performing content-based search through video materials.

On Habr, you can already find guides on creating your own virtual lecturer based on Synthesia and the Synthesia-virtual-instructor-studio GitHub repository, which automates the educational video production pipeline.

The "Generative Lecture" technology combines LLM (including GPT-5) and an instructor clone to turn a standard video lecture into an interactive experience — a student can pause the video at any time to ask a question, and the avatar will answer using materials from the entire course.

Discussion of ethical aspects is becoming an integral part of avatar implementation. Experts emphasize the need to explicitly notify students that they are interacting with a digital twin, not a live instructor. The instructor's own consent is also required to create and use their virtual copy. Legal mechanisms for this are still being formed.

The following are applied in the profile building segment:

There are also Russian developments that are gradually gaining momentum. GigaChat and YandexGPT are already used for generating lecture texts and automatic content creation, while the "Teacher Assistant" service adds feedback and analytics functionality. In terms of video generation quality, domestic alternatives still lag behind Western ones, but given import substitution requirements, their development is being accelerated.

Comparative analysis of Russian and foreign experience

If you systematize Russian and foreign experience, you can identify several key differences:

General trends coincide, they strive for personalization, increased engagement, and reducing teachers' workload.

Universal standards for exchanging profiles between educational platforms are already being developed, and avatars are moving towards full conversational interactivity. Russia holds an active position in this movement, and the experience of our universities and schools can serve as a benchmark for many countries.