CampusMind AI: A Multi-Model AI framework for Academic Management and Student Employability Analysis

This paper presented the complete design, implementation, and evaluation of an AI-Driven Student Performance and Soft Skill Evaluation System — a unified web-based platform that bridges the critical gap between academic record management and AI-powered communication skill development in engineering education. The system successfully integrates a full-featured dual-role academic portal (teacher and student) with a seven-stage AI video analysis pipeline and an LLM-powered resume analyzer. The teacher portal enables comprehensive academic management including class registration, assignment distribution with file attachment, attendance and marks recording, behavioral observation logging, and real-time submission notifications. The student portal provides filtered assignment visibility, file-based submission, AI soft skill analysis, resume review, and notification reception. The AI Soft Skill Analyzer employs OpenAI Whisper for robust speech transcription, a custom Python NLP module for vocabulary diversity and filler word quantification, MediaPipe FaceMesh for iris-position-based eye contact measurement, and Librosa for prosodic feature extraction. These four data streams are combined through a weighted multi-modal scoring framework to produce six competency scores — grammar, vocabulary, eye contact, fluency, filler words, and confidence — and an overall soft skill score. A three-tier feedback generation system (Ollama → Anthropic API → rule-based) ensures actionable coaching feedback is always generated regardless of infrastructure availability. Experimental results demonstrate end-to-end processing of 2-5 minute video recordings in 38-90 seconds on CPU hardware, producing reliable competency scores with a mean overall score of 73.1 across tested samples. The dual-database PostgreSQL architecture ensures strict data isolation between teacher and student roles, and the modular microservices design enables independent scaling of the AI engine. The system represents a replicable, open-architecture contribution to AI-augmented educational assessment. By embedding soft skill evaluation directly within the institutional academic workflow rather than as a separate tool, it reduces adoption friction and creates a natural incentive structure for student engagement. Future work targeting longitudinal tracking, LMS integration, and multilingual NLP support will further extend the platform's impact on holistic student development in engineering education.

1.                Radford, A., et al. (2022). Robust Speech Recognition via Large-Scale Weak Supervision. OpenAI. arXiv:2212.04356. 2.                Honnibal, M., & Montani, I. (2017). spaCy 2: Natural Language Understanding with Bloom Embeddings, CNNs and Incremental Parsing. 3.                Lugaresi, C., et al. (2019). MediaPipe: A Framework for Building Perception Pipelines. Google Research. arXiv:1906.08172. 4.                McFee, B., et al. (2015). librosa: Audio and Music Signal Analysis in Python. Proceedings of SciPy 2015. 5.                Crossley, S. A. (2020). Linguistic Features in Writing Quality and Development. Journal of Writing Research, 11(3), 415–443. 6.                Burstein, J., Chodorow, M., & Leacock, C. (2003). Criterion Online Essay Evaluation. Proceedings of IAAI 2003. 7.                Krejtz, K., et al. (2020). Eye Tracking Cognitive Load Using Pupil Diameter and Saccade Parameters. Scientific Reports, 8, 18092. 8.                LinkedIn Talent Solutions. (2022). Global Talent Trends: The Rise of the Whole Person. 9.                Express.js Documentation – Node.js web application framework. 10.             PostgreSQL Documentation – Official PostgreSQL database documentation.