"One brain. For the real world."
Early-stage robot software startup building a hardware-agnostic deterministic execution engine that teaches humanoids skills from one human demonstration.
Lili-o is an early-stage robot software company developing what it calls a hardware-agnostic deterministic execution engine for commercial humanoid robots. The product is meant to act as a software brain that sits above the robot's own hardware and turns high-level task intent into bounded physical control steps, so that the same skills can run on humanoids from different manufacturers. The company targets factories, warehouses and printed circuit board preparation facilities, and describes high-mix, speed-critical production work as its focus. It invites hardware OEMs, system integrators and facility operators to run pilots and offers a sandbox pilot for testing.
The company's technical position, set out in its May 2026 blog post on deterministic control, is that industrial deployments cannot rely only on probabilistic vision-language-action models. Lili-o instead combines classical control pipelines with learning-based components and uses bounded symbolic control logic, and it states that its engine runs on a megabyte-scale edge CPU footprint without heavy onboard GPUs. Its One-Shot Method, published in April 2026, teaches a robot a new skill from a single human demonstration of under 60 seconds captured with multi-modal sensing. A symbolic pipeline extracts where the end-effector makes and breaks contact, how force is applied, the approach trajectories and the key spatial relationships, and converts them into a bounded execution policy that tolerates different object poses without retraining a neural model. Skill primitives are encoded as contact geometry rather than joint angles and are transferred between robot bodies through a hardware abstraction layer; a symbolic execution engine then chains primitives into longer tasks and handles bounded failure recovery. The company says it is still early and has not published success-rate or customer results.
The founding team listed on the website is Nicolas Metzke (CEO), Ludovic Maitre (CTO), Jiseong Oh (CRO), Quentin Munch (Chief Scientist) and Antoine Forest (Head of Mechatronics). The website places the founders in South Korea, while Dealroom lists the company as launched in 2025 in Seoul with headquarters in Ile-de-France, France, and a team of 2 to 10 people. Dealroom records a seed round led by Antler in June 2026. Partners named on the website are Antler, NVIDIA, WhatsLab, Metaba and the French cobot maker Niryo, and the company reports a memorandum of understanding with WhatsLab.
Primary type & automation activities this supplier delivers:
Robot-agnostic symbolic execution layer for humanoids that chains contact-geometry skill primitives with bounded failure recovery on an edge CPU; offered as sandbox pilot.
Teaches a robot a new skill from one human demonstration under 60 seconds by extracting contact, force and trajectory constraints into a bounded execution policy.
Contact Lili-o Inc.
WEBSITE
https://www.lili-o.comHEADQUARTERS
Germany
Company Facts
Founded
2025
Primary Role
Software/Algorithm
Company Size
-
Primary Region
Europe
Annual Sales
-
Funding Stage
Early-Stage VC
Funding Total
-