Research

My research sits at the intersection of computer science and the learning sciences: I build computational models of how people learn, and I design learning environments that put those models to work. Three threads run through this work.

Learning as Graph Traversal

Dissertation · LAK26 & ICMI 2026 Doctoral Consortia

My dissertation asks what it means to learn a richly interconnected domain. Quizbowl — a competitive academic game with hundreds of thousands of questions spanning the canon — offers an unusually clean window into that question: every question rewards recognizing an answer from progressively easier clues, so every buzz is a timestamped trace of what a player knew and how deeply they knew it. I model learners' trajectories as walks over a concept graph, treating learning as movement through a knowledge space rather than the accumulation of isolated facts.

This framing turns pedagogical questions into graph problems. What does an efficient path through a domain look like? Can a curriculum be generated by planning routes through the graph? And what do thousands of players' traversals reveal about the structure of the canon itself? Early pieces of this work appear at the LAK26 and ICMI 2026 doctoral consortia, alongside talks at CoDEx 2026 and the Symposium on AI, Education, and the Learning Sciences.

Structured Hackathons & AI Education at Scale

AIED 2025 · CSCL 2026

I came to this thread from both sides of the room: first as the Grand Prize winner of SuperAI Engineer, Thailand's national AI talent program, and then as a lecturer and teaching assistant with the Artificial Intelligence Association of Thailand. Hackathons are usually treated as one-off events; my work asks what happens when they become the backbone of a curriculum.

At AIED 2025 I studied a structured hackathon curriculum that put 175 participants through seven weeks of Kaggle-style challenges woven into an intensive program of online and onsite sessions, finding sizeable learning gains — and that hackathon performance moderately predicts formal assessment outcomes. At CSCL 2026 I push the design question further: if one hack week is pedagogically effective, why not ten?

Embodied & Playful Computing

SIGCSE TS 2025 · TIILT Lab

With the TIILT Lab I design computing experiences that move. At SIGCSE TS 2025, with A. Quiterio, I examined a curriculum blending sports and computer science for 5th graders — so that young learners meet computational ideas through embodied, playful activity rather than at a desk. The throughline with my other work is the same conviction: the environments we learn in are designable objects, and getting their structure right is as much a scientific problem as an act of craft.