The Presidential Academy has developed the Russian pioneering university neuro-assistant based on generative artificial intelligence (AI) to answer questions from applicants and parents via a Telegram chat. Since the start of the admissions campaign, the bot has responded to more than 11,000 requests from over 4,000 unique users.
Named after the Academy, the neuro-assistant’s main functionality covers advice on the document submission procedure, admission rules, comparison of educational programs and passing scores, recommendations for admission to the bachelor’s degree and specialist’s degree programs. It can answer over 40,000 questions and is integrated with official university documents to find more accurate and faster answers.
The admission is an important nervous-consuming time and the Presidential Academy strives to make this process as convenient, simple and open as possible. Academiy can answer all the most critical and hot questions in a couple of seconds, provide links to regulatory documents and advice on the main areas of the Academy’s activities, such as science, international cooperation, education, etc. If a question is non-standard or requires the involvement of a specialist, Academiy will send you the contact details of the employee who can help.
Vladimir Kolodkin, Vice-Rector of the Presidential Academy:
“Every year, more than 100,000 applicants contact the Presidential Academy’s admissions committee with a wide variety of questions. To provide high-quality information and support to our applicants and their parents, we offer hotlines, and admissions officers also answer questions about admission details in the university’s official groups and channels on social media. However, we understand very well that sometimes applicants need not a fully-fledged consultation, but they rather need to be able to navigate in the diversity of the Academy’s opportunities. It was for these purposes that we have created our Academiy. It helps quickly navigate through bachelor’s and specialist’s degree programs and compare them with each other. It will answer questions about admission rules and send contact details and working hours of the admissions committee. And we sincerely believe that with that project choosing programs and entering the Presidential Academy will be even more comfortable.”
The most frequently asked questions include the major in the diploma, jobs for graduates, and how academic Olympiads scores can affect admission opportunities. Some do not fully understand how the assistant works, so they ask more than one question at once and even number the questions in order. In this case, the system will certainly only respond to the top questions in the list. The Academy’s IT engineers study all incoming questions to make the neuro-assistant even more user-friendly.
Academiy is a standalone microservice and is designed to automate and process user requests via the Telegram chat. Its primary goal is to provide accurate understanding, classification and solution of user requests using state-of-the-art approaches to natural language processing and machine learning. The neural assistant is based on open-source LLM and RAG technology, so it does not respond in a template manner, but it analyzes user requests and preferences.
It is integrated with an analytical dashboard that displays admissions campaign data in real time.
Pavel Klepinin, Vice-Rector of the Presidential Academy, Head of the Digital Block:
“All stages of development, from analytics to API integration, were completed by the Academy team, which guarantees the product’s high quality, security and ease of implementation into the university’s infrastructure.”
As the developers note, every day the neuro-assistant automatically uploads new information into the database management system, ensuring regular additional training without human intervention, which makes the system adaptive and capable of quickly responding to changes in user needs.
They have also created a Metabase-based analytical dashboard to monitor performance and response quality and to analyze user requests, thus quickly identifying and tackling malfunctions and to improve the system as a result.
Sergey Bolovtsov, Director of the Research Center, Presidential Academy:
“The admissions campaign is always a critical issue for universities. Our team at the Research Center for Empowering Admissions Officers at the Presidential Academy has developed a unique intelligent solution that will help answer a wide range of questions from applicants and parents. Now it continues to learn, but during the admission campaign it will be one of its most accomplished experts and will become a loyal assistant on the path to choosing a program or submitting documents; it will use in dialogue mode to explain nuances and will add relevant links to the website sections in addition.”
The Academy intends to build a cycle of continuous improvement for the neuro-assistant. As soon as intelligent solutions are implemented, the profile of user interaction with the system gets immediately updated to cover new types of questions and the wish to learn “everything at all” in dealing with the bot. For example, one of the questions that the bot received demonstrated the user’s desire to learn how to bypass anti-plagiarism checks right away.
“We are interested in two tracks, namely to be able to identify users and to provide personalized services. We are going to scale the neuro-assistant to a wider range of tasks – to advice on continuing education, master’s degree and postgraduate programs and the life of the Academy as a whole,” adds Pavel Klepinin.
Applied technologies:
- Programming language: Python 3.12 (with use of Poetry)
- Backend: FastAPI
- Embedders: E5, BM25
- Embedding options: Milvus, TEI
- Database: MongoDB
- Analytics: Metabase
- User interface: Telegram
- Infrastructure: Docker, REST API
- ML infrastructure and data processing: LangChain, LangGraph, PyTorch, pandas, numpy
