Area
Hardware-aware AI systems
I work on AI features that respect latency, memory, sensing, and deployment constraints instead of assuming infinite compute.
Portfolio ポートフォリオ
I build across embedded hardware, industrial sensing, simulation, and applied AI/ML.
My background spans factory-floor AIoT, embedded sensing, full-stack product prototypes, and technical simulations. I focus on turning real-world constraints into useful software behavior.
About 概要
Engineer bridging hardware constraints, industrial data, and product-oriented AI systems.
AIoT and embedded systems engineer with three years of industry experience in Japan, now completing a B.Eng. in Electrical, Electronics and Information Engineering at Nagaoka University of Technology.
I started in Japanese manufacturing, building data collection and quality-analysis systems that had to work with real sensors, real operators, and real production limits. Since then I have continued moving across embedded systems, applied AI, and technical product design.
My recent work spans automotive hardware architecture exposure, manufacturing AI consulting, full-stack product experiments like Rouvis, and interactive simulation projects that make complex technical ideas easier to test and explain.
Focus Areas 専門領域
The overlap I care about is where physical systems, data, and software all affect each other.
Area
I work on AI features that respect latency, memory, sensing, and deployment constraints instead of assuming infinite compute.
Area
I have built sensor-integrated tooling and analysis workflows for production environments where instrumentation and reliability matter.
Area
I move between hardware context, data pipelines, and user-facing software when the problem cannot be solved cleanly from one layer alone.
Area
I use simulations and real-time environments to make technical systems easier to test, explain, and iterate on.
Selected Work 主なプロジェクト
A curated shortlist of public work that best represents how I think and build.
AI planning workspace for new farmers in Japan, combining maps, scheduling, analytics, and conversational guidance.
Problem New farmers need support across crop planning, local knowledge, and day-to-day operating decisions without juggling separate tools.
Built Built a desktop-first Next.js application that brings map-based planning, analytics, knowledge access, and AI chat into one workflow, and the project was later featured by ETSUZAN as one of its selected teams.
Student directory and profile management app for an electrical and information engineering program.
Problem Department profiles and student information were difficult to manage, search, and update in one secure place.
Built Built a full-stack university directory with Google auth, admin workflows, Prisma-backed data management, and responsive UI.
Interactive ceramic 3-point bending simulations for teaching, rapid prototyping, and early-stage fracture studies.
Problem Brittle-fracture experiments and solver comparisons are usually split across scripts, validation tooling, and separate interactive demos.
Built Combined Python validation scripts, config-driven FEM workflows, and a Godot 4 demo with an optional native C++ backend.
ESP32-based current transformer monitoring pipeline with Firebase-backed remote data access.
Problem Current sensing experiments needed a low-cost way to capture measurement data and inspect it remotely without a heavy infrastructure stack.
Built Implemented an ESP32 and Firebase workflow for CT-clamp measurements, using cloud storage and a simple web-facing monitoring path.
Torque measurement visualization and analysis tool for industrial inspection data and pass/fail evaluation.
Problem In-house torque measurement data needed faster comparison, feature extraction, and evaluation without manually checking raw CSV files.
Built Built a Python application that visualizes normal, filtered, and FFT views, applies threshold-based checks, and generates downloadable analysis outputs.
Co-op ocean extraction game prototype with authoritative multiplayer systems and a shared builder loop.
Problem Multiplayer extraction games often bolt progression and teamwork clarity on too late, which makes systems feel disconnected.
Built Built a Godot prototype with a shared hangar builder, replicated boat systems, runtime damage, and evolving reward loops.
Trajectory これまでの流れ
The through-line is consistent: build useful systems at the boundary between constraints and capability.
Aug 2025 to Sep 2025
Worked on automotive E/E architecture and ADAS hardware platform customer-interface work in a hybrid internship setting.
Jul 2025 to Aug 2025
Implemented Fusion RAG system work for the agentic AI chat product OPTiM AIRES.
May 2023 to Apr 2025
Built sensor data pipelines, quality analytics, and machine-learning workflows for production environments.
Apr 2022 to Apr 2023
Worked on electric-vehicle motor-core manufacturing process development before moving into industrial IoT engineering.
2025 to present
Delivered image-analysis and vibration-classification tools for manufacturing clients after leaving full-time industry work.
2025 to 2027
Currently studying electrical, electronics, and information engineering in a coding theory and data-sequence lab while extending research and independent product work.
Contact 連絡先
If you are hiring or building in embedded AI, industrial software, or technically demanding product environments, I am easy to reach.
I prefer concise technical conversations with clear problem statements. GitHub is the best place to see public work; email is best for collaboration, hiring, or project discussions.
Public repositories and recent work
Professional profile and career history
Best for collaboration and hiring inquiries
Manufacturing and hardware background