We are Moza Amer Al-Subhi and Maisa Hamad Al-Tobi, supervised by Rolando Jr Lontok. Our project is the Smart Academic Advising System for UTAS Nizwa.
Advising today is manual and repetitive: transcripts are read by hand, the same rules are explained again and again, and students at risk are noticed late.
Our proposed solution is one advising platform for four user roles. A rule engine calculates every academic result, Google Gemini explains the results in plain language, and the advisor makes the final decision.
CIMS remains the official university system. The platform supports advising; it does not register students.
Our five objectives: a secure role-based system; transcripts with GPA, standing, progress and alerts; rule checks and recommendations; plan review, appointments and notifications; and versioned data with audit.
The first two objectives are completed in this version. Objectives three to five are upcoming.
Comment one: outdated literature. We added 2022 to 2024 studies on advising, explainability and conversational AI, in Chapter 2.
Comment two: the diagrams. We redrew the framework, the ER diagram with keys and cardinalities, and the flowchart with standard symbols.
Comment three: the role of AI. Now the rules decide, Gemini only explains approved results, and the advisor keeps the final judgement.
Comment four: security. We specified role-based access, scoped data, row level security and audit logs; the demo already enforces roles and sessions. We also moved to Next.js, TypeScript and Supabase.
Comment five: the research gap. Ours is integrated, explainable advising for UTAS with human review and data governance.
Comment six: methodology. Each objective is now mapped to its data, analysis and evidence, with the sample limits stated.
The official transcript and versioned rules go in, deterministic checks run, and results come out: GPA, standing, alerts and plans.
People review the outputs and govern the rules; AI explains, it never decides.
Four roles use one Next.js application. The transcript upload is the only bridge from official records.
Behind the server checks, the decision engine calculates and the AI advisor explains its output only.
Everything is stored with versioned rules and audit records, in Supabase through Prisma.
Identity and roles: each user has one profile, and many students link to one advisor. Actions go to the audit log.
Students have enrollments, plans and appointments; each term points to the policy version that was applied.
Every session starts with sign-in. Unknown or disabled accounts stop here.
Each role then follows its own lane. A student's plan passes the rules before the advisor decides, and every decision is audited.
Objective one: sign-in with four roles and scoped access are done. Allowed and denied flows are verified; hosted deployment is next.
Objective two: upload, GPA, standing and progress are done; the trend chart and what-if are next. All 26 automated tests pass.
One sign-in page, four destinations: each role lands on its own home.
Every page checks the role on the server. A student who types the advisor address is sent back to their own overview.
A branch admin creates students and advisors and assigns each student an advisor.
The server filters accounts by role and branch: the branch admin sees three, the super admin sees all five.
Each account has its own hashed password, and a password reset ends every old session immediately.
We read the official transcript and check every grade against the policy. A bad file gives a clear error and nothing is saved.
The student sees a preview first: name, GPA, standing and every course. Nothing is saved yet.
Confirm saves only if the preview is the student's own, still fresh and unchanged; otherwise the old record stays.
Grades become points from the published policy, Article 79.
Semester GPA is credit-weighted: 15 points over 9 credits, 1.67, in both semesters.
DEMO103 was repeated, and the most recent attempt counts: 30 over 15, a cumulative 2.00.
The latest semester is below 2.00, so the standing is probation, with a 12 credit limit.
DEMO104 is a fail, so 12 of 15 imported credits are passed.
And this is the code: the points come from the policy, never from the code itself.
The student's overview shows the same numbers, each with its rule code.
Alerts say what to do next, and a note is honest that progress covers imported coursework only.
The advisor sees only their assigned students; an unassigned student is not available.
This is the system architecture by implementation status: green is completed, amber is partly completed, for example running locally instead of on Supabase, and grey is upcoming.
Our path to the final submission: Supabase and deployment, rule checks and recommendations, Gemini explanations, plans and appointments, and versioned configuration.
Thank you.