R&D LAB · CYBERINFRASTRUCTURE

Secure AI & Educational Chatbots — research to deployment

We design, pilot, and evaluate AI-driven chatbots and secure cyberinfrastructure that support research, teaching, and operational decision-making. Our work emphasizes reproducibility, ethics, and robust security.

Selected Projects

Laboratory projects focus on conversational agents for education, secure model deployment, data governance, and visualization dashboards for institutional research.

EduCyber: Educational Chatbot Pilot

Pilot deployment of an AI-driven conversational agent to collect student feedback, support instruction, and generate analytic insights for faculty. Designed for ethical deployment with human oversight and anti-misuse safeguards.

  • Conversational surveys & guided tutoring flows
  • FERPA-aware data capture & secure cloud repository
  • Lightweight on-premise / cloud inference options

Secure Inference & Model Governance

Practical approaches for model registry, reproducible pipelines, and secure serving that preserves privacy while enabling auditability and reproducible R&D.

Tech stack: model registries, RBAC, logging, TLS hardened endpoints, drift monitoring.

Campus Dashboards & Visualization

Interactive dashboards to visualize engagement metrics, sentiment analysis, and course-level trends for faculty and administrators.

Deliverables include Plotly/Chart.js dashboards and exportable analytic reports for program evaluation.

Human-in-the-loop Evaluation

Assessment protocols and rubrics to validate chatbot behavior, minimize bias, and ensure pedagogical alignment.

Methodology & Data Practices

Secure Data Capture

We use encrypted repositories, least-privilege access, and metadata standards so data can be shared safely with collaborators while preserving student privacy.

Key components

  • FERPA-aware storage (encrypted buckets)
  • Metadata standards for interoperability
  • Audit logs & provenance tracking

Chatbot Design Principles

Guided-assistance flows that scaffold student learning instead of giving direct answers, combined with guardrails against misuse.

  • Progressive hinting & mastery checks
  • Transparent model outputs with citations
  • Human-in-the-loop fallback for edge cases

Sample Insights (demo data)

These sample visualizations illustrate the type of trends we produce from chatbot interactions and engagement logs.

Exportable reports • Reproducible analyses

Engagement: Weekly Active Users (sample)

Interpretation: rising usage over the pilot period with peaks after targeted faculty prompts.

Sentiment: Chatbot Feedback (sample)

Interpretation: sentiment distribution from conversational surveys (positive/neutral/negative).

Project Timeline (sample plan)

Phase 1 — Preparation

Design chatbot flows, IRB planning, faculty onboarding.

Phase 2 — Pilot

Deploy to selected courses, collect conversational & LMS metrics.

Phase 3 — Analysis & Reporting

NLP sentiment analysis, dashboarding, and final recommendations.

Team & Collaborators

profile
Dr. Hong Jiang
PI — AI & Educational Technology
profile
Research Engineers
Modeling, systems, and security
profile
Student Researchers
Undergraduate & graduate assistants

Resources & Publications

Technical Reports

Summaries of pilot findings, reproducible notebooks, and model evaluation notes.

Open Source

Code examples, conversation flows, and anonymized datasets for re-use under appropriate agreements.

Contact & Collaboration