About Atom
We build the robot Japan needs now.
Atom was founded in 2024 in Aomi, Tokyo by engineers from industrial robotics research who had watched Japan's factory labour gap accumulate year on year — and who knew from the inside that fixed-arm automation was not going to close it.
The bet is geometric: the tasks that remain on the manual side of the automation boundary — irregular objects, multi-step pick-place-rotate sequences, operations at non-standard orientations — are precisely the tasks that require a body shaped like the workstation was designed for. The workstation is built for a person. The robot should be too.
Founding Story
Tokyo, 2024
Shunsuke Aoki spent five years studying whole-body motion planning for humanoid systems at an industrial robotics research institution in Tokyo. The gap he kept observing was not in the research — bipedal locomotion, compliant manipulation, imitation learning had all made genuine progress. The gap was in who was turning that research into hardware designed for Japan's actual factory floors, with Japan's actual safety regulations, and Japan's actual operator expectations.
Japan's manufacturing labour shortage is not a temporary tightness. Working-age population has been contracting since the 1990s. The jobs that remain manual are the jobs that fixed-arm automation was architecturally unable to take — irregular geometries, non-standard sequences, tasks where the environment changes faster than the reprogramming cycle. These are not the hard cases. They are the common case.
Atom assembled in Aomi, Tokyo in 2024 with a specific mandate: build a system that a plant operations director can actually evaluate, trial, and deploy — not a research platform seeking a use case, but hardware specified against tasks that are costing manufacturers labour every shift.
Principles
Three decisions we made early
01
No general-purpose claims
Atom is designed for factory floor and logistics yard tasks in Japanese industrial environments. We do not claim it will work in a hospital, a construction site, or a household. Scoping the problem to Japan's manufacturing sector made the design constraints tractable. Claiming universality would have made them not.
02
Describe limits before deployment
Every briefing includes a frank account of what the robot cannot currently do — force resolution limits, payload ceiling, tasks where human operators still outperform the system. We do not discover these limits during a customer's trial deployment. We describe them first, so the trial is scoped to what we can actually deliver.
03
Japanese regulatory environment as a starting constraint
Industrial Safety and Health Act requirements, Labour Standards Act provisions for human-robot workcells, and the operating protocols that Japanese facility operators actually use — these are the constraints the design was specified against from day one, not adapted to afterward.
The Team
Six engineers. One problem.
Compliant actuator design, whole-body motion control, multi-modal perception, RTOS-based embedded systems, and manufacturing operations discipline — the full stack required to take a humanoid from a research concept to a machine that can work a factory shift.
Shunsuke Aoki
Co-Founder & CEO
Whole-body motion planning research, Tokyo. Founded Atom in 2024 when the gap between research and factory deployment became an engineering problem he could close.
Keiko Tanabe
Head of Mechanical Engineering
Series elastic actuator design for precision automation. Leads Atom's 29-DOF kinematic chain and joint architecture.
Hiroshi Watanabe
Lead Perception Engineer
Structured-light depth, RGB classification, and fingertip tactile fused into Atom's scene graph perception stack. Specialist in grasp planning under mutual occlusion.