Improving Capstone Team Outcomes through Dynamic Skill Matching and Preference Alignment

Synthesis: Pardi et al. (2026) introduce a three-stage methodology for dynamic student team formation that integrates student preferences with project skill requirements. Students complete a survey, an LLM analyzes project descriptions to extract skills, and a dynamic assignment algorithm matches students to projects. Preliminary evaluations show higher skill coverage and preference satisfaction than random or manual assignment, overcoming limitations of CATME Team-Maker — addressing a core problem in team-based engineering and computing education.

Team-based projects are a cornerstone of engineering and computing courses, but unstructured team formation often leads to poor project outcomes due to misaligned student interests and inadequate skill coverage. This paper introduces a novel, three-stage methodology for creating effective student teams by integrating student preferences with project skill requirements. Students complete a survey, an LLM analyzes project descriptions to extract skills, and a dynamic assignment algorithm matches students to projects. Preliminary evaluations show higher skill coverage and preference satisfaction compared to random or manual assignment, overcoming limitations of CATME Team-Maker.

• LLM-driven team formation outperforms CATME Team-Maker for skill coverage and preference alignment in capstone courses.

What this means for practice

• Instructors. Run the three-stage pipeline (survey, LLM skill extraction, dynamic assignment) instead of forming capstone teams by hand: it fulfilled 98.4% (Fall 2023) and 91.9% (Spring 2024) of project skills against 90.4% and 89.6% for manual assignment, which took over 20 hours of instructor effort.
• Instructors. Accept the small preference trade-off that automation brings, and say so to students: average preference toward the assigned project was 87.2 and 86.3 with the algorithm versus 88.7 and 87.6 for manual assignment, while random assignment managed only 58.2 and 64.0.
• Instructors. Collect skill self-ratings and ranked project preferences as first-class inputs, then set expectations about project scope early: students who prioritized an "interesting topic" sometimes reported lower satisfaction when reality did not match.
• Designers. Review the LLM-extracted project skills before assignment runs: the authors note that LLM-generated project skills still require manual verification.

Limitations

• Evaluation covers two semesters of one program — the UC Merced computer science capstone — with Fall 2023 at 16 projects and 68 students and Spring 2024 at 22 unique projects, some of them shared by two teams.
• The satisfaction evidence is a brief survey of only 22 previous capstone students, self-reported on seven Likert and multi-select items.
• The algorithm depends on self-reported skills and preferences, leaving it vulnerable to misrepresentation, and the preference weight α requires manual tuning.
• All skills are treated equally, with fulfillment resting on a single student's intermediate-level rating, so projects with many required skills can be favored over those with rare but critical skills.

Citation

Brandon Pardi, Garret Castro, Michael Pisman, Avash Adhikari, Santosh Chandrasekhar (2026). Improving Capstone Team Outcomes through Dynamic Skill Matching and Preference Alignment. CSCI 2025 (12th International Conference on Computational Science and Computational Intelligence).