Rafkat Abaev, a UK higher education student systems engineer, recommends data-first governance that fixes definitions, validation, and integration contracts before you add any AI layer.
Universities are moving fast on AI. EDUCAUSE data shows the speed: the share of institutions with AI-related acceptable use policies rose from 23% in 2024 to 39% in 2025. However, when they try to connect AI to real campus systems, many teams discover that the hard part is fragmented data sources and under-documented institutional processes.
AI cannot scale in higher education if the underlying student data is inconsistent, non-interoperable, or impossible to audit, Rafkat Abaev argues. Since 2012, he has focused on UK higher education digital platforms and the student lifecycle systems across major UK universities, such as King’s College London, University College London, Queen Mary University of London, University of the Arts London, and others. He builds what he calls data-first governance, where data quality and consistency logic come before interfaces and features. For the establishments and the industry, his work consistently delivers three outcomes: fewer reporting errors and less operational risk, better accessibility with lower support overhead, and faster, more dependable processing of applications. Across institutions such as UCL, Queen Mary, UAL, and UWE Bristol alone, the student environments tied to this work serve well over 140,000 students based on published university figures, which gives that governance work real institutional scale. He also holds a patent for an explainable risk prediction and recommendation system designed for education analytics, which connects the governance grade foundation to trustworthy automation.
Using Rafkat’s data-first approach as a lens, we show how to turn fragmented student records into a trustworthy backbone for analytics, early-warning systems, and responsible AI.
Interoperability is the bottleneck
Most institutions run a patchwork of platforms across student records, learning, identity, advising, assessment, and communications. Data can move between systems, but without interoperability and shared meaning it moves slowly, expensively, and inconsistently. According to OECD, when tools are not interoperable, linkage and sharing become error-prone and resource-consuming, which limits the value institutions can extract from their digital ecosystem.
“This is why integration standards exist. Think of them as common connectors that let tools share basic information safely and predictably. For example, one standard helps a learning app plug into a university platform without custom code for every campus. Another helps systems exchange rosters and enrollment data so the right students get access to the right courses”, Rafkat comments.
That focus on standards and integration maturity also shows up outside campus work. At the AITEX Summit in 2025, Rafkat served on the Expert Board, where he, as an expert in the field, assessed projects through a formal evaluation framework for architecture, technical execution, and real-world applicability rather than demo polish alone.
At the same time, global AI guidance is pushing institutions to show how decisions are made, what data was used, and who is responsible when something goes wrong. The broader AI governance world is moving the same way: document what the system does, put checks around it, and make it auditable. In other words, you can demo AI on imperfect data, but you cannot run it as a reliable campus service unless the data and integrations are under control.
The student record is where accountability lives
Many AI conversations in education start with the learning management system. Accountability is anchored elsewhere. Enrollment status, progression decisions, awards, eligibility, and official reporting depend on the student record and the processes around it.
That is why the student information system becomes the hidden constraint. It is where what is true gets defined. It is also where mistakes are hardest to unwind, because errors propagate into downstream decisions and external obligations.
Rafkat’s work is concentrated in Student Information System (SITS) Vision and its portal layer eVision, the kind of system that runs the student journey from recruitment and admissions through support, retention, and outcomes. His relevance to the current AI moment is not that he shipped a chatbot. It is that he repeatedly engineered the data and workflow conditions that make automated decision support credible. This is less a one-off technical fix than a reusable operating logic: the same approach to validation, workflow structure, and reporting control was applied across several university environments and carried forward in later modernization work.
Audit grade data is engineered into workflows
In regulated environments, data quality is a requirement with hard deadlines. Most countries have official student data submissions that affect public reporting, compliance, and sometimes funding (in the UK, this is done via the Higher Education Statistics Agency — HESA). Working as a Student Information System, or SITS, developer with universities such as University College London, Queen Mary University of London, and Birmingham City University, Rafkat Abaev took on the reporting layer where inconsistencies typically surface at the last moment. In the HESA stream specifically, he redesigned the reporting scenarios inside the student records system, translated reporting requirements into automated rules, added validation and control checks that flag gaps and contradictions before submission, and fixed mismatches between academic records and what the reporting framework expects. The work was done under tight submission windows and high-pressure periods like Clearing and major system upgrades, where an error is not just a defect but a potential compliance risk. The stated outcome was fewer reporting errors, stronger compliance confidence, and less manual rework for administrative teams.
If an institution cannot consistently define a student’s status or outcome across processes and reporting, predictive models will amplify inconsistency. Audit grade data discipline is what makes later explanations and accountability possible.
Rafkat is a Senior Member of IEEE, a grade reserved for professionals with at least ten years in practice and significant performance over at least five of those years. The jury board of the international community considered his approach valuable and repeatable. His idea that auditability is an engineering requirement that has to be built into rules, validation, and traceable states is considered valuable by engineering communities.
Where data lineage becomes non-negotiable
Interoperability issues surface first when universities connect their core student records system to national and government services, especially during peak admissions cycles. At the University of the West of England, Rafkat led an upgrade of the student information system and modernised the integrations that move applicant and enrollment data between the university and external platforms, including the UK’s national admissions service and government education interfaces. In his case descriptions, the impact is operational: fewer manual handoffs, faster and more reliable processing, and better synchronisation of records across systems.
Many early AI use cases target admissions and recruitment. Triage, personalisation, yield forecasting, and early risk flags depend on clean mappings and traceable data lineage. Without that, teams stay stuck in pilots.
If workflows break, automation follows
Portal modernisation is often treated as a user interface exercise. In practice, it fails when the underlying workflows break under peak usage. Abaev led the conversion of student portal tasks connected to the university student records system (SITS/eVision) into responsive, mobile-friendly formats at Queen Mary University of London, and delivered similar responsive migrations plus admissions and enrollment workflow enhancements at University College London. In his case descriptions, the impact is operational: better accessibility, fewer support tickets, and fewer issues when students use the portal on mobile devices.
He also worked on high-load digital services at the University of the Arts London, including the academic portal, student messaging, assessment and progression workflows, and dynamic notifications. Another case focuses on stabilising an admissions peak season process at London Metropolitan University, where application volume surges and systems are under maximum stress, improving reliability when the institution can least afford outages.
“If the system cannot behave predictably during peak load, automation and AI features will not behave predictably either,” he sums up.
Explainability and monitoring are not optional add-ons. They are requirements that only work when the underlying data and workflows are engineered for consistency and traceability first.
AI is forcing universities to get more disciplined. Clearer data definitions, cleaner integrations between systems, and governance that can stand up to scrutiny are becoming basic requirements, not best practices. This will decide which AI use cases become stable parts of student support and academic operations, and which stay stuck in pilots. Rafkat Abaev’s work sits inside that shift. It shows what university digital transformation looks like when it moves beyond interface upgrades and into traceable data, interoperable systems, and governance that can stand up to scrutiny. In that sense, his contribution is not rhetorical but operational: he helps make the kind of AI adoption the sector is now moving toward actually workable.