Academic Integrity · AI Era
Every university is in the same fight: AI has made content-detection a losing game. UniPortal takes a different position. We capture how work is actually made, and we verify that the student can teach what they submitted. Evidence replaces accusation.
Editorial thesis
The platform replaces detector theatre with a reviewable chain of evidence: how work was made, how it changed, and whether the student can defend it.
Live
Comprehension checks fire during writing, not after
Verified
Authorship evidence captured across multiple sessions
Pending
Teaching Test scheduled after final submission
Trace
Declared quotations and revision depth remain visible
Built for institutions · 12 data tables · 3 role systems · live demo
scroll ↓
The Context
Universities spent a decade building content-detection infrastructure. Then transformer models made authorship unknowable from text alone.
Every month, detectors get less reliable. False positives accuse honest students. False negatives pass through work the student cannot defend.
0%+
false positive rates across leading detectors*
“Detection cannot keep pace with generation.”
This is not a tooling problem. It is a framing problem. The question stopped being did this student use AI years ago. The useful question is can this student explain what they submitted. UniPortal is built around that question.
UniPortal · Observed 2024—2026
The Mechanism
Scroll to see it work ↓
The Verify Framework
Stage 01 / 06
Live comprehension · Timed interruption · Adaptive follow-up · Evidence captured during writing
Verified compose session
Session 1 · 00:04:32 active
Stakeholder theory proposes that organizations have obligations to multiple parties beyond shareholders...
This is the mechanism. Live at uniportal.com.au.
The SDOT Framework
Phase 01
Students research, read, gather evidence, and build understanding before drafting. The platform captures engagement with sources — not clicks, not time-on-page, but intentional annotation and reference.
SYSTEM VERIFICATION · Source engagement · Research notes · Evidence of preparation
Journal article
Annotated evidence on assessment validity
Lecture notes
Student note set · 14 references captured
Source map
Argument scaffold linked to references
Inside the Product
COMPOSE · TRACE · VERIFY · PROVE · MATCH · INSIGHT
Compose
A TipTap-based editor captures sessions, timing, revision depth, and paste declarations without interrupting the student. Accessibility-first, accommodation-aware, and built to make authorship evidence feel native instead of punitive.
Compose Workspace
TrueLearn Draft Environment
Session 1 of 5
12 min active
Verify understanding, not just sources.
Live Evidence
Paste Event
Monitoring paste declarations in real time.
Trace
Trace engine · Sample submission
7 sessions over 10 days
4h 32min total active time
1,247 words typed · 104 quoted
3.4 edit events per paragraph
Verify
UniPortal interrupts writing with random, timed checks on content the student just produced. Sixty seconds to explain. Scores accumulate into the integrity score, with adaptive frequency and accommodation-aware pacing.
SYSTEM VERIFICATION · Real-time · Adaptive frequency · Accommodation-aware
Verify engine · Live during writing
Comprehension check incoming · 00:09
UniPortal interrupts writing with random, timed checks on content the student just produced.
Adaptive signal
Sixty seconds to explain. Scores accumulate into the integrity score. Strong students see fewer checks, struggling students see more.
SYSTEM VERIFICATION · Real-time · Adaptive frequency · Accommodation-aware
Prove
After submission, Claude-generated questions across simplify, justify, counter, extend, process, and connect probe whether the student can defend their own work — the final check on top of the live comprehension checks captured during writing.
Teaching Test
Question 03 of 05 · Justify
Why does UniPortal prefer comprehension verification over trying to label whether text was written by AI?
Scoring Model
Simulated Result
92 / 100
Match
After three submissions, the platform builds a baseline across vocabulary, structure, rhythm, and voice. New work is compared against that history with enough nuance to distinguish growth from anomaly.
Student baseline
Students tend to write with shorter clauses, moderate vocabulary, and a direct argumentative rhythm. Paragraphs average 142 words. First-person references appear sparingly.
Current submission
The current draft shows stronger academic vocabulary and denser sentence structure, but paragraph flow and transition choices remain recognisably aligned with the prior baseline.
Instructor heatmap
A.B.
92
Low
E.A.
78
Review
K.N.
34
Flagged
L.H.
88
Low
S.T.
64
Review
P.M.
95
Low
Insight
Heatmaps surface risk at a glance. Detailed evidence reports turn a 30-minute review into a 3-minute decision. Approvals, flags, and resubmissions are logged for institutional defensibility.
Platform Coverage
The live platform already behaves like a university operating layer: a front door for coursework and balances, a role-aware demo and login flow, a persistent student workspace, and adjacent support surfaces that make TrueLearn feel embedded in campus operations instead of bolted on.
Campus front door
A single entry point for coursework, fees, schedules, integrity workflows, academic support, and progress tracking.
Courses
Enrol & track progress
Fees
Payments & balances
Integrity
TrueLearn verification
Grades
Reports & analytics
Schedule
Timetables & events
AI Assistant
Smart chatbot support
Role-aware entry
Lecturer and student pathways are explicit from the first interaction, so each person enters the product with the right permissions, context, and evidence model.
Lecturer
Dr. Maya Chen
Lecturer account
Student
Avery Morgan
Student
Jordan Patel
Student
Sofia Bennett
Navigation Model
The left-hand information architecture is already doing real product work. Main academic navigation, the dedicated TrueLearn writing surface, management tasks like fees and schedules, and academic support tools all sit in a single persistent shell.
Main surfaces cover dashboard, courses, assignments, submissions, grades, and library access.
TrueLearn Write is treated as its own product layer rather than a buried assignment utility.
Management surfaces keep fees, schedules, and integrity operations within the same session.
Support surfaces include the AI academic assistant and notifications, not just back-office admin links.
Bernard Adjei-Yeboah
Online
Main
TrueLearn
Management
Academic Support
Student Workspace
The dashboard in the product is not a placeholder. It connects academic performance, unit load, pending work, upcoming deadlines, schedule, fee status, integrity scoring, and AI study support into one session. That breadth matters because it positions UniPortal as a campus workflow layer, not just an assessment tool.
Here's what's happening with your academic journey today.
Current GPA
6.67
Units Enrolled
3
Pending Tasks
5
Integrity Score
92%
Upcoming deadlines
View all
Final Report · ICT6001
Due 2025-12-18
Research Presentation
Due 2025-12-22
Peer Review Submission
Due 2025-12-15
Today's schedule
4 events
Applied Project Lecture
Room 301
Group Meeting
Library Study Room B
Supervisor Consultation
Office 204
AI Workshop
Computer Lab 2
Unit progress
Trimester 2
TrueLearn score
92 / 100
Fee status
Paid in Full
Trimester invoice cleared. Statement and receipt history are available on demand.
AI study assistant
Powered by GPT-5
Instant help with assignments, research, and study materials — without leaving the student workflow.
What this adds
Performance cards tie academic momentum to operational context: GPA, units, pending tasks, and the integrity score sit in the same top layer.
The deadlines and schedule surfaces show that UniPortal already behaves like a student planning environment, not merely a submission portal.
Fee status and statement access extend the platform into administrative workflows that students actually revisit during the term.
The TrueLearn breakdown makes the scoring model legible by exposing authorship, comprehension, and consistency as separate evidence pillars.
The AI Study Assistant, explicitly presented as GPT-5-powered, gives the platform a visible model layer that supports learning rather than just policing it.
The Artifact
Three submissions. Three different stories.
Instructor Report
low risk
92
Typed
1,247
Quoted
104
Undeclared
0
Recommendation
Approve
Session Timeline
Teaching Test
Comprehension score, review notes, and flag context change with each scenario.
92
Seven distributed sessions, declared quotations, and a strong Teaching Test result support approval.
Flags
Action
Approve
Logged actions, notes, and follow-up requests become part of the institutional record.
Live sample — data is illustrative, not from a real student
For Students
UniPortal is not surveillance. It is the opposite — an environment that captures evidence you did the work, so your grade can stand on something real. Accommodations-aware, accessibility-first, designed for the person doing the writing.
For Instructors
Heatmaps show where attention is needed. Evidence reports replace guesswork with data. When you flag a submission, you have something to show.
For Institutions
GDPR, FERPA, and Australian Privacy Act compliance built in. WCAG 2.1 AA accessibility. Audit logs. Role-based governance. Single sign-on when you need it. A platform an academic board can say yes to.
The Technical Position
UniPortal is designed to meet the practical standards universities actually care about: privacy, accessibility, governance, and a technical foundation that does not collapse the moment scrutiny begins.
Security & Privacy
Compliance
Infrastructure
About the Builder
Bernard Adjei-Yeboah
Builder of UniPortal · Applied AI, product systems, and university workflow design
Institution
Asia Pacific International College
Program
Master of Information Technology
Specialisation
Artificial Intelligence
Project
UniPortal
Bernard Adjei-Yeboah is the builder behind UniPortal, a product shaped around a practical question universities now face in the AI era: how to verify genuine understanding without relying on brittle detection theatre.
The platform was developed in the context of Bernard's Master of Information Technology studies at Asia Pacific International College (APIC), where he specialised in Artificial Intelligence. That academic foundation is central to the product itself: UniPortal is designed to show how AI can be applied with rigor, fairness, and institutional responsibility.
APIC's academic environment provided the setting in which this work could be taken seriously as more than a classroom interface. UniPortal reflects an effort to translate postgraduate research, systems thinking, and product craft into something a university can actually evaluate, discuss, and potentially deploy.
Master of IT
Developed within a Master of Information Technology journey grounded in applied systems thinking.
AI Specialisation
Focused on artificial intelligence as a practical discipline for real institutional products and workflows.
APIC Context
Asia Pacific International College provided the academic setting, challenge, and credibility behind the work.
Built with Intent
UniPortal was shaped as a serious product demonstration for student success, academic integrity, and university operations.
UniPortal is built for institutions that want to verify learning, not prove its absence. The platform is live. Come see how it works.
Verify understanding, not just sources.