IntellCollab: A Machine Learning-Driven Recommender System for Academic-Industry Project Collaboration

Collaboration between academia and industry faces systemic challenges including misaligned objectives, communication barriers and inefficient matching of expertise to project requirements. This paper presents IntellCollab, a web-based platform that addresses these challenges using an enhanced content-based recommender system. The system employs term frequency-inverse document frequency (TF-IDF) vectorisation with cosine similarity for initial matching, augmented with Word2Vec embeddings for semantic understanding and evaluated against stateof-the-art baselines. Developed using a hybrid agile-waterfall methodology, IntellCollab incorporates role-based authentication, skill-based querying and a dynamic dashboard interface. We evaluate the system on a curated dataset of 3,100 academic and industrial projects, achieving 88.0% accuracy and an F1-score of 0.89 in project recommendation. Ablation studies confirm the benefit of our hybrid approach and statistical significance testing (p < 0.01) demonstrates consistent gains over baseline methods. We detail the system's technical architecture, evaluation methodology and error analysis and discuss its potential to enhance knowledge transfer and research-industry partnerships.