AspireAI
Smart Career Guidance
AI-Powered Career Guidance System for Sri Lanka
AspireAI
About the SystemCareer guidance challenges such as uninformed stream selection, Z-score uncertainty, and lack of career visibility significantly impact Sri Lankan students' academic performance, yet traditional advisory methods often fail to address these effectively. This research proposes an integrated AI platform leveraging machine learning, time series forecasting, collaborative filtering, and NLP to identify optimal educational pathways. The platform analyses academic grades for stream selection, Z-score data for university admission forecasting, employment market data for career pathways, and speech features for soft-skill assessment. Trained on 7,094 student records from 15 schools across 3 provinces and evaluated with 450 learners over 6 months, AspireAI delivers 86.89% stream accuracy, 3.2% Z-score MAPE, and an 82.4 SUS usability score β classified as Excellent.
Research Foundation
Explore the literature, problem, gap, objectives, methodology and technologies behind AspireAI.
Literature Survey
Several AI tools have been developed to assist with educational recommendations. Existing systems using Random Forest achieved 89.5% accuracy while XGBoost ensemble approaches reached 91.3%, but lacked interactive curriculum validation. Z-score forecasting for university admission was largely unexplored in developing-country contexts. Studies on RAG-based LLM grounding show it minimises hallucination in educational settings, making it ideal for quiz generation and course advising. Collaborative filtering methods have been applied in e-commerce recommendation but rarely in structured career pathway contexts for students with non-traditional qualifications.
Research Problem
Sri Lankan students face critical educational transitions without integrated, data-driven guidance. Only 15β20% of approximately 300,000 annual Advanced Level candidates secure state university admission. The remaining majority must navigate fragmented, poorly documented alternatives β private universities, vocational courses, and overseas options β without systematic support. The lack of a holistic platform connecting O/L stream selection β A/L performance β Z-score forecasting β employment readiness creates a significant equity gap across income groups, geographic regions, and school quality tiers.
Research Gap
A review of 47 peer-reviewed publications revealed no existing platform integrating all of the following:
- Predictive stream selection with curriculum-aligned validation quizzes
- Z-score forecasting with ROI and scholarship estimation
- Career pathway visualisation for non-traditional (non-university) learners
- Objective, automated soft-skill assessment using speech analysis
- All designed specifically for Sri Lanka's bilingual, multi-stream, UGC-regulated ecosystem
International platforms (Khan Academy, Naviance) address some of these individually but are not contextualised for Sri Lanka's unique Z-score system, district-based university allocation, or local labour market conditions.
Research Objectives
- Design and implement a hybrid XGBoost + RAG pipeline that recommends optimal A/L academic streams based on O/L results, subject interests, and career goals, with interactive curriculum-aligned quiz validation.
- Develop an ARIMAβProphet weighted ensemble model that forecasts district-specific university admission Z-scores and estimates ROI for different degree programs.
- Build a collaborative filtering + content-based hybrid system that generates personalised career pathways for learners entering non-traditional post-secondary routes (TVEC, NIBM, private, overseas).
- Create an ASR + NLP soft-skill assessment pipeline that transcribes student speech samples, extracts fluency and coherence features, and classifies proficiency with adaptive recommendations.
System Methodology
AspireAI is built as a four-module integrated platform using microservices architecture with FastAPI backends and a React + Material-UI frontend.
Module 1 Β· Stream Recommendation
XGBoost multi-class classifier (Ξ·=0.05, depth=6, n=200) with RAG pipeline (FAISS + all-MiniLM-L6-v2). Trained on 7,094 records from 8 schools across 3 provinces. Oversampled with SMOTE.
Module 2 Β· Z-Score Forecasting
ARIMAβProphet weighted ensemble (w=0.6/0.4) on 3,847 course-district-year UGC records. Weighted scoring: Z-score compatibility (50%), subject alignment (30%), interest match (20%).
Module 3 Β· Career Pathway Generator
Hybrid SVD collaborative filtering + cosine-similarity content filtering (Score = 0.6Β·CF + 0.4Β·CB). LightGBM ranking with Gemini narrative generation grounded in local labour market data.
Module 4 Β· Soft Skill Assessment
ASR transcription (13.2% WER) with NLP feature extraction: fluency, lexical richness, coherence, sentiment. 2-layer neural classifier (ReLU, dropout=0.3) maps to 3 proficiency levels.
Technologies Used
AspireAI integrates state-of-the-art tools across ML, NLP, backend, and frontend domains.
Technologies
System Performance
Validated on 450 learners over 6 months across four AI modules.
Project Milestones
All research assessments with dates, mark allocations, and completion status.
Project Proposal (Presentation + Proposal Report)
Initial project scoping, problem definition, objectives, and proposal presentation to the panel. Project Charter signed by supervisors.
Progress Presentation β 1
Literature review, data collection across 15 schools, architecture design, and initial XGBoost model prototypes demonstrated to the panel.
Research Paper Publication
IEEE conference paper submitted and accepted, documenting the AspireAI methodology, datasets, and preliminary results across all four modules.
Progress Presentation β 2
All four modules integrated and demonstrated. Interim: 91.8% stream accuracy, 4.6% MAPE. Full React + FastAPI system integration shown.
Final Reports (Thesis)
Longitudinal study with 450 learners over 6 months. Final: 86.89% accuracy, 82.4 SUS score, 23% increase in informed enrollment decisions.
Final Presentation β’ Viva
Final oral examination and project defence. Each member defends their module contribution with a live demonstration.
Website Development
Research portfolio website development and deployment to the SLIIT CDAP portal. Built with HTML / CSS / JavaScript.
Project Documents
All research documents produced throughout the project lifecycle. Click to download.
Project Charter
The document gives information regarding the statement of scope, objectives overview, an outline of scope, approximate schedule and people participating in the project.
Download PDFProject Proposal
The document contains details like goals, objectives, important dates, milestones and requirements needed to start and complete the research project.
Download AllResearch Paper
A research paper containing literature review, methodology, analysis, interpretation, and argument based on in-depth independent research. Published in IEEE.
Download PDFFinal Thesis (Main)
The document contains the proposed solution to the research question, finalised after six months of longitudinal evaluation with 450 learners across Sri Lanka.
Coming SoonIndividual Documents (Γ4)
Contains individual per-member component thesis documents β one for each of the four research modules developed by the AspireAI team.
Coming SoonCheck List Documents
The document describes the progress of the project within specific time periods and compares it against the project plan milestone completion checklist.
Coming SoonBusiness Plan
The document outlines the commercialisation roadmap, revenue model, and go-to-market strategy for AspireAI as a scalable EdTech platform in Sri Lanka.
Coming SoonPresentation Slides
Download slides from all past and upcoming presentations.
Meet the Team
The researchers behind AspireAI, under the guidance of SLIIT Faculty of Computing.
Ahamed A L I
Specialising in Information Technology
Ahmed M A A
Specialising in Information Technology
Bandara R M M K T
Specialising in Information Technology
Areeb Aflah N
Specialising in Information Technology
Jenny Krishara
Faculty of Computing, SLIIT
Poorna Panduwawala
Faculty of Computing, SLIIT
Get in Touch
Got a technical issue? Want to send feedback? Need details about collaboration or our research? Let us know.
We'd love to hear from you
For inquiries about the AspireAI research project, collaboration opportunities, or general questions, feel free to reach out through the form or via the details below.
+94 11 254 4802 (Faculty of Computing)