Chalmers AI Sommarprojekt 2025 - Team Smart Uni

Project Overview

During summer 2025, Team Smart Uni conducted an extensive exploration of AI-driven solutions for the university world. Over a ten-week period, we conducted approximately 15 in-depth interviews with staff from across the Chalmers organization (and a few from external universities) for four main projects addressing challenges in education, student support, research networks, and alumni relations.

Research Process and Methodology

Our data collection consisted of recorded interviews that were transcribed via the Sana platform, integrated with Microsoft Teams. Documentation occurred continuously through weekly blog posts in the open Teams group "Chalmers Framåt".

We conducted interviews with:

• Study advisors - to understand student support processes
• Pedagogical experts - to map learning and technology integration
• Fundraising and alumni activities - for strategic insight into alumni engagement
• CRM specialists - for data analysis and segmentation
• Research administrators - for research networks and bibliographic data
• Department management - for organizational perspective
• International collaboration partners - for global perspectives on AI in education

Central Research Questions

We explored four fundamental questions:

• Can we build a digital twin of a student to identify those who may need support in advance?
• How can AI be used for personalized learning?
• Can we create more individualized communication with Chalmers' 70,000 alumni?
• How can direct and indirect connections be mapped in a research network?

Ethical Considerations

Throughout the process, we considered:

• Student privacy and right to privacy
• Voluntary participation and opt-in principles
• Transparency in AI systems' decision processes
• Quality assurance of information sources

Project Solutions

StudyBuddy - The Future of Learning

"The future of learning is most likely highly personalized, as each student has different knowledge, experiences, abilities, and study method preferences."

StudyBuddy evolved from a simple chatbot idea to a sophisticated knowledge graph-driven learning assistant. The system addresses the fundamental problem that many students struggle with scientific texts and course literature, leading to superficial understanding.

Key Features:

• Knowledge Graph - Maps relationships between concepts, courses, and prerequisites
• Personal Knowledge Meter - Continuously tracks mastery of each concept through multiple signals
• Smart Repetition - Uses Hermann Ebbinghaus forgetting curve and spaced repetition algorithms
• Adaptive Teaching Methods - Employs evidence-based pedagogical approaches like fading, retrieval practice, and dual coding
• Optimal Difficulty - Implements the 85% rule for "desirable difficulty"

Student.Twin - Predictive Student Support

A digital twin system designed to identify students who may need support before problems become critical. The system analyzes various data points to predict which students might struggle and proactively offer assistance.

ResearchBook - Intelligent Research Network Mapping

An AI-powered system that maps direct and indirect connections in research networks, helping researchers discover expertise across different institutions and facilitating collaboration.

Smart Alumni - Personalized Alumni Engagement

A comprehensive platform for managing alumni relationships through:

• Intelligent Segmentation - Automatic clustering of alumni into meaningful groups
• Predictive Models - Identifying alumni most likely to donate, volunteer, or engage
• Email Intelligence - Optimizing timing and content for maximum engagement
• Media Monitoring - Automatic detection of alumni achievements in news

Key Insights from Stakeholder Interviews

• Students rarely participate in events and information meetings because they don't feel the need "right then"
• Traditional Canvas structures vary dramatically between courses, making data-driven analysis difficult
• Alumni communication is too generic - "the majority of alumni are not super engaged, but with the right approach you can activate the long tail"
• The research landscape is fragmented and difficult to overview - researchers often don't know what expertise exists at other institutions
• Current research databases are static and lack flexibility for asking complex questions or visualizing collaborations

Technical Implementation

Each solution was built using modern AI/ML technologies including:

• Knowledge Graphs for representing complex relationships
• Machine Learning Algorithms for predictive modeling
• Natural Language Processing for content analysis and generation
• Spaced Repetition Algorithms for optimal learning schedules
• Clustering Algorithms for intelligent segmentation

Results and Impact

The project demonstrated that it's technically possible to:

• Automate target group segmentation with high precision
• Predict engagement and donations based on open data
• Optimize communication timing for maximum impact
• Identify successful alumni in real-time via media monitoring
• Create personalized learning experiences that adapt to individual student needs