Nanjie (Jimmy) Rao

An AI tutor teaches from what reaches it about a student. I study where that information should come from — asking, teaching, or the records students already produce.

Topics

Research

  • interest-based learningTeaching through what each learner already cares about: my dissertation’s starting point.
  • crowd-sourced sensingSensors, residents and posts must agree before a flood event is confirmed.
  • teacher support toolsThree teacher-facing papers: IEEE ICALT 2021, HCII 2022, IEEE ICALT 2023.
  • recommender explanationsHow explanation types and user motivations shape recommender use (AAAI HCOMP 2020).
  • HCIPh.D., Human-Centered Computing, University of Florida, December 2026.
  • mixed methodsQualitative coding with a second coder, plus GEE, bootstrap and jackknife.
  • coding schemes28 codes in four categories measure what a tutor learns about an interest.
  • interest phasesMy Phase construct: where an interest is in its development, situational or individual.
  • LLM tutorsWhat an AI tutor must know about a student, and where it comes from.
  • asking vs. reading recordsAsk the learner, or read the records they already produce? I compared both.
  • browsing historyHigher precision than posts in my ICALT 2026 study: 69.8% of its interests confirmed.
  • social media postsCarried more grounded detail per confirmed interest than browsing history.
  • Accuracy · Depth · Richness · PhaseThe four constructs my dissertation uses to judge interest identification.
  • grounded vs. inferredChecking each claim a model makes about a person against their own material.
  • privacy by designSelf-curation, sensitive-content filtering and non-retention, built into my study protocol.
  • urban digital twinsA design framework and a flood prototype; oral presentation at DTRUE 2026.

Building and sensing

  • hand enactmentHow writers return to recorded hand enactments (ICIDS 2026, N = 24).
  • idea reuptakeHands as Narrators (ICIDS 2026) characterizes idea reuptake from hand enactment.
  • embodied learningHands, puppets, avatars and AR glasses in storytelling and everyday learning.
  • digital puppetsA low-cost digital puppet system for an NSF storytelling project (HCII 2021).
  • motion trackingYOLO vision plus IMU sensors driving puppets in Unity in real time.
  • eye trackingWebcam gaze (WebGazer.js) and Tobii Pro Glasses 3 in my builds.
  • AR smart glassesTwo Vuzix Blade apps, compared in a 25-person study (IEEE VR 2020).
  • telepresence robotOn-site leader of Team Limitless, ANA Avatar XPRIZE semifinalist (2021).
  • activity recognitionBreadboarding and soldering filmed with Tobii glasses; GPT-4o summarized activities from frames.
  • Unity 3DReal-time puppet animation driven by external tracking.
  • KinectUndergraduate thesis: a humanoid robot imitating human actions.
  • FPGAPart of my hardware toolkit, with PCB CAD and microcontrollers.
  • solderingHands-on hardware: soldering, PCB CAD and 3D printing.
  • PCB designPCB CAD, part of my hardware toolkit.
  • wearablesA wearable baby thermometer, later commercialized and sold in the U.S.
  • 3D printing3D-printed puppets tracked by vision and IMUs.

Teaching and life

Research

An AI system that adapts to a person must first know something about them, by asking or by reading records the person already produced. My research asks which to rely on, how accurate and well grounded the result is, and what it costs in privacy.

Research overview

Recent

  1. Upcoming oral presentation at DTRUE 2026 (UF Jacksonville campus, Nov 19–20, 2026): our design framework and flood prototype for crowd-sourced urban digital twins. Civic digital twins
  2. Published academic-revision-checklist, an open checklist for revising academic English: 66 rules and a 16-item quick list from five weeks and more than 60 revision rounds of my dissertation. It also works as a Claude Code skill. Academic revision checklist
  3. Passed my oral defense. Dissertation: Toward Scalable Interest-Based Learning: Comparing Data Sources for AI-Based Automated Identification of Students’ Personal Interests. Interest identification

Latest video

Lucky levels up at Okito Taekwondo class - form combination test ·

Play video: Lucky levels up at Okito Taekwondo class - form combination test (YouTube)

More on YouTube

On GitHub

More on github.com/JimmyRaoUF.