The founder of the international education company 4 Prep Academy on a transparent, step-by-step scoring system, adapting marketing to local markets, an algorithm for redirecting applications after a visa refusal, and AI-avatar teachers.
EdTech has become an inseparable part of academic consulting: digital services are now used at every stage of university admissions, from analyzing an applicant’s chances to preparing them for interviews with admissions committees. Abdiaziz Abdishukurov, founder of the consulting company 4Prep Academy, which helps students gain admission to leading universities in the US, Canada, Europe, and China, develops tools based on predictive analytics and generative models to automate international recruitment. The predictive system he created, 4Prep AI, uses machine learning to match applicants with the universities where they have the best chance of admission. We spoke about how artificial intelligence is changing the work of educational agencies and opening a path to US universities for talented young people from around the world.
International education recruitment is becoming less and less predictable: visa rules keep changing, competition for students is growing, and the choice of university still often depends on the experience of a particular consultant. How did you manage to translate that expertise into the 4Prep AI service?
The conventional approach is limited by the human factor and a lack of verifiable statistics. A consultant simply cannot keep constantly changing requirements from thousands of foreign institutions, their scholarship caps, and past years’ visa histories in their head at all times. 4Prep AI solves this by converting accumulated experience into an algorithmic model. Instead of a limited list of familiar universities and a rough estimate of costs, the program compares a student’s profile against hundreds of programs and produces a precise calculation of their probability of receiving a scholarship. We digitized data from 10,000 cases collected over six years of work, which rules out guesswork. The system also continuously updates its data on scholarship funds and visa statistics, adapting its recommendations to constantly changing admissions rules. As a result, the applicant gets a mathematically grounded admissions strategy with minimized risk.
One common criticism of predictive products is a lack of transparency in how decisions are made. To address that, you developed the Educational Pathway Intelligence Framework, which underlies 4Prep AI. How exactly does this scoring system work?
Our tool works like an internal digital calculator built on mathematically matching a user’s metrics against our database. First, the system gathers precise data about the candidate through an interactive questionnaire-based profiling process: academic performance, language test scores, the family’s financial limits, and visa history, including any prior visa refusals from the US or European countries. In the second stage, the algorithm compares these inputs against university requirements and filters accordingly. For example, if a candidate has strong athletic achievements but the family’s budget is five thousand dollars, the algorithm immediately excludes institutions without grant programs and instead selects universities where candidates with similar profiles have previously received actual funding in the range of forty to fifty thousand dollars. In the third stage, the user receives a detailed strategy: an approved list of institutions, the optimal grant amount to request, and backup options in case of visa risk.
Is getting the visa still the most vulnerable stage?
Yes, visa policy is often unpredictable, and even a candidate with excellent academic performance can be refused. So as not to leave a person in a hopeless situation, we built in a redirection algorithm. If a student is refused a US visa, our program automatically activates pre-booked backup options at European universities in Italy, Germany, or Latvia. We identify these institutions during the initial profile analysis stage. European schools offer solid English-language programs, and their visa requirements are more transparent. This preserves the student’s educational path, and it also helps our company protect its commercial performance: the client switches to an alternative destination while continuing to work with us.
How did you technically organize support for applicants, given that every university has its own requirements and every student has their own set of documents and their own admissions timeline?
Off-the-shelf CRM systems work well for quick transactional sales, but they aren’t built for the complex logic of academic consulting, which unfolds over months. An applicant has to complete dozens of tasks: writing a personal statement, taking international exams, collecting recommendation letters. Our internal tracking portal, built on a modern tech stack, combines a reliable PostgreSQL database with a Next.js front end through the Prisma ORM. That gives us full control over every action. The system automatically logs progress: it tracks document preparation status, certificate uploads, and application deadlines. Our Vercel and Supabase infrastructure ensures fast interface performance and secure data storage. For us, that has translated into a significant speed-up in processing requests, and for clients, into greater transparency — the ability to see the status of their case in real time.
When a consulting company’s staff work across different continents, there’s a risk of misaligned actions, delayed responses to clients, or even missed application deadlines. You built a virtual COO, AI-COO Aziz, to manage your Tashkent team from Los Angeles and avoid exactly those breakdowns. What tasks did you hand off to it?
As a business scales, coordination between departments eats up an enormous amount of resources. Automation lets us take routine work off managers’ plates. Our lead-scoring architecture, built on ManyChat and the Claude API, analyzes prospective clients’ inquiries on social media and classifies their academic level and financial capacity right from the first point of contact. That lets the sales team focus on the most promising leads. To manage the Tashkent office from California, I built a virtual COO, Aziz, on the Claude API and ElevenLabs. The bot is integrated into our corporate messenger. It independently analyzes employees’ written reports, checks task deadlines, sends reminders about unfinished work, and reads out summaries in a synthesized voice. This has cut administrative costs and sped up task completion without needing to expand the management team.
I know you see a lot of EdTech’s future tied to the growth of synthetic media. You regularly take part in the ETC Synthetic Media Summit at the University of Southern California, and you have your own project, AI World, featuring photorealistic AI avatars. Are users ready to accept a digital teacher as a genuine conversation partner?
One-on-one lessons are constrained by cost and a teacher’s schedule, while recorded courses don’t allow for real dialogue. In AI World, we combine HeyGen avatars with language models and ElevenLabs speech synthesis. A teacher like this can answer questions and explain material around the clock, in multiple languages. We don’t see this technology as a replacement for human teachers — its purpose is to make personalized learning accessible in regions where qualified specialists are in short supply. The platform is aimed primarily at markets in Asia, Latin America, and Africa. For us, access to education isn’t a privilege — it’s a right that technology helps make real.
You build technologies that automate expert work, yet you continue to mentor at Aitex and Hackathon Raptors, and you recently became a judge for the Armenia Digital Awards. How does evaluating other people’s projects help you improve 4Prep AI?
When I evaluate a project, I look first at practical applicability, scalability, and data security. Judging and mentoring let me compare dozens of architectural approaches, spot new ways of integrating models, and notice common vulnerabilities. We use these observations to develop 4Prep AI further: refining algorithm logic, testing hypotheses, and adopting proven engineering approaches. Young specialists get feedback on launching tech products, and I get a broader base of material to draw on for developing our own solutions.
