Work with me
Have a robot that works in simulation but not reliably in the real world?
I help robotics teams turn perception, navigation, and multimodal AI ideas into testable, deployable systems.
ROS 2 migration & performance
Port ROS 1 stacks to ROS 2, adopt lifecycle nodes, profile and remove bottlenecks. Case: ~50x segmentation speed-up.
Perception pipelines
LiDAR/camera perception, segmentation, traversability and costmaps for mobile and field robots.
VLA & imitation-learning setup
Teleoperation rigs, LeRobot data collection, and fine-tuning open VLA models on your robot.
Edge deployment
Jetson and Raspberry Pi bring-up, camera and audio integration, CI running on device.
Acoustic sensing R&D
Microphone-array systems, impulse-response measurement, and audio machine learning.
Prototypes & interactive robots
Vision plus voice interaction and bilingual (EN/JA) human-robot interfaces.
Process
Scope → prototype → deliver
01
Scope
A short call to define the failure mode, the target metric, and what 'done' looks like on your hardware.
02
Prototype
A minimal, measurable slice on real data or the real robot — no month-long silence.
03
Deliver
Production-quality code, tests, documentation, and a benchmark you can re-run without me.
Headline numbers
Evidence, not adjectives
~50x
Segmentation speed-up
0.2 → 10 FPS in ROS 2
16-ch
Microphone array
OccluNet acoustic rig
40
Teleop episodes
Bimanual VLA dataset
6+ yrs
Robotics experience
Japan-based, field-tested