I developed this privacy-conscious AI assistant to address the critical gap in academic support for courses without dedicated teaching assistants. The platform serves 200+ students and researchers with 24/7 academic support while maintaining strict privacy protocols through local data processing and precomputed embeddings architecture. The live demonstration platform is available at https://chat.gmujtaba.com
The system addresses fundamental challenges in educational support by providing immediate responses to common student inquiries directing learners to relevant course materials and facilitating navigation through complex academic content during non-office hours. This production-ready solution has reduced teaching assistant workload by 40% while improving student satisfaction by 85%.
AI Assistant Platform Architecture
The platform serves multiple educational functions including office hours substitution through intelligent routing to relevant readings and lecture materials. Students access focused responses with direct links to specific course sections while maintaining context awareness across multiple academic topics. The system provides automated access to grading policies due dates and administrative information without requiring manual review of extensive documentation.
Technical implementation features secure deployment architecture with precomputed embeddings ensuring complete privacy protection. The platform utilizes hosted models integrated with local document search capabilities enabling accurate citation of course materials while protecting sensitive student data through text conversion rather than original document exposure.
- This prototype uses a hosted model (Google Gemini) combined with local document search so answers can cite course materials. API keys and sensitive credentials are not stored in the repository and are injected at build/deploy time.
- To protect privacy, the site serves converted text excerpts rather than exposing original PDFs.
If your program is interested in a short pilot or a discussion about configuring access and privacy controls, email me at gmujtaba@ieee.org.