JAENContact

RESEARCHER×ENGINEERcreate the principlescale it to society

Making XR a safe,
everyday infrastructure.

As a Ph.D. student, I research how to measure the state of people using XR so they can collaborate safely across the real and the virtual. Through Nexreal LLC, I also take on software development for companies, including large-scale SaaS.

My goal is the “Interverse”, where the real and the virtual merge so people can work together beyond the limits of place, body and time.

Kodai Matsudaira松平 幸大
  • Ph.D. student, Tokai University (Hamamoto Lab)
  • Software Engineer, Nexreal LLC
Contact
01 / Research

Measuring people to make
collaboration in XR safe.

For XR to become everyday infrastructure, we need to measure human limits such as sickness, cognitive load and walking risk, and design collaborative spaces that span the real and the virtual around them. These are my two themes.

THEME 01

Human-centered safety infrastructure

Integrating physiological, behavioral and subjective measures during XR use to quantify sickness, cognitive load and walking risk, and building state-estimation models and safety evaluation protocols.

THEME 02

Collaboration across the real and the virtual

Clarifying design principles for co-presence and task performance in asymmetric collaboration that mixes CAVE, HMD and AR.

Research projects

4 PROJECTS
R-01 · THEME 01

Objective assessment of AR sickness with physiological signals

AR sickness has so far been measured only by questionnaires after the experience, missing how it develops and the signs before people notice it. We record gaze, ECG and EEG together during AR use to find which physiological responses track the strength of sickness. From there, we aim to build objective indicators that catch sickness before symptoms appear.

2 related talks
R-02 · THEME 01

Safe information display on AR glasses while walking

Looking at information on AR glasses while walking can draw attention away and slow reactions to the surroundings. We measure reaction time, gait and autonomic responses together when content is shown at the center or the edge of the view, and quantify how placement changes the load. From the results, we aim to derive design guidelines for AR displays that stay safe while walking.

In progress
R-03 · THEME 02

Image quality design for 360° images in an immersive CAVE

How good a 360° image looks when projected in a CAVE differs from person to person and does not match objective quality metrics. We test whether the three types of quality preference we found are specific to the immersive CAVE or stable traits of individuals, by comparing the same images in the CAVE and on a regular screen. From there, we aim to build image quality guidelines suited to immersive environments.

2 related talks
R-04 · Applied research · SICK K.K.

Stamped character recognition combining 2D and 3D features

Characters stamped on parts have little contrast with their background and look different under different lighting, so images alone cannot read them reliably. We test how far recognition improves with preprocessing that combines image texture (2D) and surface shape from a 3D sensor (3D). We aim for reliable automatic reading of stamped characters for industrial inspection.

2 related talks
02 / Engineering

From concept
to production.

Through Nexreal LLC I take on whichever stages a project needs, from requirements to operations, and carry products beyond PoC all the way to production.

AI features for an education SaaS

Adopted by municipalities

As part of a development team working with PMs and QA, built AI features that support students’ learning and teachers’ instruction in a domestic education SaaS used by millions.

  1. Reqs
  2. Design
  3. Detail
  4. Build
  5. Test
  6. Ops
  • Designed a shared client for multiple LLM providers and optimized cost with prompt caching and cost tracking
  • Designed and built anonymization that masks student and teacher data before it reaches the LLM

TypeScript · Next.js · AWS · Amazon Aurora · GraphQL · Python · FastAPI

Recruiting workflow automation

Workload cut by over half

Automated candidate touchpoints in recruiting, from calls and SMS to scheduling and assigning staff.

  1. Reqs
  2. Design
  3. Detail
  4. Build
  5. Test
  6. Ops
  • Built scheduling that finds free slots across several recruiters’ calendars and prevents double booking
  • On booking, automatically registers the event, sends confirmation email and notifies the team chat

Express · React · Supabase · Google Workspace API · GCP · Docker

Management dashboard

Real-time

Consolidated scattered sales data into one real-time view of revenue, conversions, win rate and LTV, for both management decisions and individual performance.

  1. Reqs
  2. Design
  3. Detail
  4. Build
  5. Test
  6. Ops
  • Implemented row-level incremental sync from spreadsheets to the DB, plus full resync for recovery
  • Re-architected from event-driven scripts to a containerized API server as requirements changed

TypeScript · Express · Google Apps Script · Supabase · Retool · GCP · Docker

Q&A chatbot

Used by several clients

A bot that answers questions in Slack based on internal documents.

  1. Reqs
  2. Design
  3. Detail
  4. Build
  5. Test
  6. Ops
  • Built a mechanism that collects experts’ answers and keeps growing the knowledge base
  • Added image attachments so users can ask questions with images, not just text

TypeScript · Slack Bolt · Dify · Supabase · GCP · Docker

03 / Publications

Publications & talks

Full list on researchmap
20264 items
  1. 2026.08

    生理指標の経時変化特性を用いたAR酔い感受性の予測

    2026 IEEJ Electronics, Information and Systems Society Conference

    TalkDomestic
  2. 2026.07

    A Hybrid Preprocessing Approach for Stamped Character Recognition using 2D Texture and 3D Shape Features

    2026 International Symposium on Flexible Automation (ISFA 2026)

    TalkPeer-reviewedIntl.
  3. 2026.03

    生理指標を用いたAR酔いの客観的評価に関する検討

    2026 Annual Meeting of the IEEJ

    TalkDomestic
  4. 2026.03

    没入型立体環境における360度画像の投影 ―第2報―

    IPSJ 88th National Convention

    TalkDomestic
20253 items
  1. 2025.09

    Mirroring Agent：自己の感情状態を反映するエージェントの設計とその対話がもたらす自己認識への影響

    The 30th VRSJ Annual Conference

    TalkDomestic
  2. 2025.07

    3D-OCR AIとDX

    JSST 9th NBC Technical Committee Meeting

    TalkDomestic
  3. 2025.03

    没入型立体環境における360度画像の投影

    IPSJ 87th National Convention

    TalkDomestic
20241 item
  1. 2024.03

    音声認識を用いた疑似会話による英語学習コンテンツの開発

    2024 Annual Meeting of the IEEJ

    TalkDomestic
04 / Articles

Articles

No articles yet.

05 / Experience & Education

Experience

WORK

  1. 2026.01 —

    Nexreal LLC

    Software Engineer — software and R&D

  2. 2025.05 — 2025.12

    NEXASPARK Inc.

    Engineering support for client projects

  3. 2024.08 — 2025.05

    RCAST, The University of Tokyo

    System development for research projects

  4. 2024.05 —

    SICK K.K.

    Applied research and development for sensors

  5. 2024.04 —

    Information Technology Center, Tokai University

    Researcher

  6. 2023.11 — 2024.03

    Digital Design Studio Inc.

    Eye-tracking support for a Tokyo-funded XR project

EDUCATION & FELLOWSHIP

  1. 2026.04 —

    JST SPRING

    Fellow

  2. 2026.04 —

    Graduate School of Science and Technology, Tokai University

    Ph.D. program (expected March 2029)

  3. 2024.04 — 2026.03

    Graduate School of Information and Telecommunication Engineering, Tokai University

    Master’s program

  4. 2020.04 — 2024.03

    Department of Information Media Technology, Tokai University

    Bachelor’s program

06 / Contact

Contact

For research collaboration, development work, talks or interviews, please reach me at the address below.

  • Research & collaboration
  • Development work (Nexreal LLC)
  • Talks & interviews
[email protected]Send an email
Contact