Yuki Aruga
Machine Learning Engineer
Musashino
Summary
Machine Learning & Robotics Engineer experienced in applying
advanced ML algorithms including — including Vision-Language Models
(VLMs) and computer-vision recognition - to robots operating in
human-centered environments. Skilled at bridging research and real-
world deployment through computer vision, LLM-based agents, and
exploratory work with reinforcement and imitation learning in robotics
PoCs. Eager to further develop expertise in learning-based control
while delivering practical innovations.
Overview
3
3
years of professional experience
2
2
Certifications
Work History
Robotics Engineer, AI Engineer
NTT Docomo
04.2022 - Current
- Designed and deployed VLM-inspired perception and attention-state estimation modules on Jetson Orin + Realsense platforms, integrating outputs into ROS2-based navigation to enhance robot safety in offices and cafeterias.
- Built a coffee delivery PoC using Temi robots, combining Slack/web ordering and autonomous navigation, achieving 1.5× sales pace during trials.
- Led development of task-planning AI agents in Habitat simulation, enabling collaborative human–robot behavior.
- Developed LLM-powered systems:
RAG enterprise Q&A app deployed before Microsoft Copilot solutions.
AR × LLM communication prototype (patent pending).
Fine-tuned in-house LLM “tsuzumi” via PEFT/LoRA and deployed LLM inference servers + webUI (Ollama + GPT-OSS + OpenWebUI). - Rescued and completed an accounting data analytics project, designing dbt/Trino pipelines and PowerBI dashboards for management KPIs
Environment & Tools: Python (PyTorch, NumPy, Pandas), C++ (ROS2 perception modules), YOLOv8, OpenCV, Jetson Orin, Realsense, Habitat-Sim, Isaac Sim, Gazebo, vLLM, HuggingFace, Docker, dbt-Trino, PowerBI, Git/GitHub/GitLab.
Education
Master of Engineering Management - Robotics
Agriculture And Technology University of Tokyo
Tokyo 04.2001 -
Skills
Programming: Python (PyTorch, Scikit-learn, Pandas), C (ROS2, perception modules)
Publications
- Autonomous Navigation of Robots Based on Human Posture for Enhanced Comfort – JSME Robotics and Mechatronics Conf., 2021
- Attention-State Recognition and Costmap-Based Robot Navigation – JSME Robotics and Mechatronics Conf., 2020
- 2 Patent applications (pending): AI × AR communication systems, generative AI technologies（特願2023-184069、特願2024-002058）
Certification
TOEIC L&R: 725
demo（sample）
- Human Attention–Aware Navigation – OpenPose + OpenCV based system integrated with ROS2 navigation
Video https://youtu.be/PLPHdxJKwZs?si=yATaapxHlF6kfomE - AR × LLM Communication Prototype – Patent-pending prototype using LLM and AR glasses
Demo https://www.youtube.com/shorts/klL47KYtjxI - Person Detection with Jetson Orin Nano + YOLOv8 – Real-time human position detection for robots
Video https://youtu.be/yq17JLQkq4U - Open-source Differential Drive Robot – Hardware & ROS2 stack for autonomous navigation
GitHub https://github.com/mikaka-robotics/raspi-rover
Timeline
TOEIC L&R: 725
02-2023
Robotics Engineer, AI Engineer
NTT Docomo
04.2022 - Current
Japan Deep Learning Association – G Test (AI Engineer Certification)
01-2022
Master of Engineering Management - Robotics
Agriculture And Technology University of Tokyo
04.2001 -