Research

I research adaptive, AI-driven learning: how a system can model what a learner knows, decide what to teach next, and explain it in natural language that stays grounded in the course.

Directions

1
Learner modelling
Estimating what a student knows from noisy answers, so the system can adapt without asking the learner to self-report.
Bayesian Knowledge TracingSpaced repetitionError taxonomies
in EduVision, LangVis
2
Reinforcement learning for pedagogy
Learning a policy that keeps each learner in the flow zone - hard enough to grow, easy enough not to quit.
PPOReward shapingA/B evaluation
in EduVision
3
Grounded conversational tutors
LLM tutors that follow a pedagogical strategy and stay factual by retrieving from course material.
Retrieval-augmented generationNeuro-symbolic controlReal-time speech
in EduVision, LangVis