COMPASS AI – Smart Competency-Based Assessment and Personalized Learning
Inspiration
Traditional education often measures student performance through marks and grades, but these numbers do not always reveal how well a student understands individual concepts. A student may score well overall while struggling with specific topics, or perform poorly without knowing exactly where improvement is needed.
This inspired us to build COMPASS AI, an intelligent learning platform that focuses on what students truly understand, where they struggle, and how they can improve through personalized, data-driven learning.
What It Does
COMPASS AI aims to transform traditional assessment into personalized, competency-based learning.
Key features include:
- AI-Powered Competency Estimation: Estimates student mastery at individual skill and topic levels.
- Skill-Gap Detection: Identifies concepts that require additional attention.
- Progress Tracking: Monitors changes in competency as students complete assessments.
- Personalized Learning: Supports targeted practice recommendations based on identified weaknesses.
- Competency Analytics: Helps students and educators understand learning progress beyond overall marks.
How We Built It
We designed COMPASS AI around a machine learning workflow that converts assessment responses into meaningful competency insights.
- Data Processing: Python, Pandas, and NumPy help clean and organize student assessment data by skill and topic.
- Competency Modeling: Bayesian Knowledge Tracing (BKT) serves as our baseline approach for estimating how student mastery evolves across successive attempts.
- Predictive Modeling: Logistic Regression and XGBoost are candidate models for evaluating student performance prediction.
- Backend Architecture: FastAPI is proposed to handle assessment processing, model integration, and application APIs.
- User Interface: React is planned for an interactive dashboard that displays competency profiles, learning gaps, and progress.
- Data Management: PostgreSQL is proposed for storing assessment records and competency profiles.
We selected the ASSISTments Skill Builder 2009–2010 dataset as the foundation for exploring student responses and skill-level learning patterns.
Challenges We Faced
One of our main challenges was moving beyond overall marks to represent student understanding at the individual skill level. Mapping assessment questions to the correct skills and interpreting repeated responses requires careful data processing and modeling.
Another challenge is distinguishing a temporary mistake from a genuine knowledge gap. We also need to ensure that competency estimates are meaningful and that recommendations provide practical guidance rather than generic suggestions.
We are addressing these challenges through structured preprocessing, a baseline knowledge-tracing approach, model evaluation, and a modular application architecture.
What We Learned
Through this project, we explored educational machine learning, student-performance analysis, knowledge tracing, and the design of AI-powered learning systems.
We learned that an effective educational AI solution requires more than a predictive model. Its results must be understandable, actionable, and useful for both students and educators. We also gained experience in connecting data processing, machine learning, backend services, and user-interface design into a unified workflow.
Impact
COMPASS AI aims to help students take ownership of their learning and enable educators to make more informed academic decisions.
By identifying learning gaps earlier and supporting targeted practice, the platform has the potential to encourage continuous improvement, personalized learning, and more meaningful competency development in higher education.
What's Next?
Our future roadmap includes:
- Implementing adaptive question selection based on estimated competency.
- Comparing BKT with supervised machine learning models.
- Developing interactive student and faculty dashboards.
- Integrating learning recommendations and progress reports.
- Exploring integration with institutional Learning Management Systems (LMS).
Our Vision
Move beyond marks. Understand mastery. Personalize every student's learning journey.
COMPASS AI envisions an educational experience in which every student can understand their strengths, identify their weaknesses, and receive clearer guidance on what to learn next.
Log in or sign up for Devpost to join the conversation.