RLWRLD
RLWRLD Inc. is a Seoul-based physical AI company building foundation models for dexterous robot manipulation, licensing RLDX-1 and running multi-year Robotics Transformation engagements with industrial enterprises in manufacturing, logistics, retail, hospitality, and airlines.
- Company typePrivate
- Founded2024
- HeadquartersSeoul, South Korea
- Headcount51–100
- GTM typeB2B
- OfferingSoftware
What RLWRLD does
RLWRLD Inc. is a Seoul-headquartered physical AI company that develops foundation models for the unsolved problem of dexterous manipulation in humanoid robots. Its flagship product, RLDX-1, is an 8.1-billion-parameter dexterity-first foundation model built on a proprietary Multi-Stream Action Transformer (MSAT) architecture that processes vision-language, action, proprioception, memory, tactile, and torque modalities through dedicated streams with joint self-attention; integrated sub-modules include a Motion Module for spatio-temporal reasoning, a Physics Module for tactile and torque signals, a Cognition Interface with 64 learnable tokens functioning as long-horizon memory, and a robot-specialized VLM fine-tuned from Qwen3-VL 8B. A three-stage training pipeline (pre-training, embodiment-specific mid-training, post-training via DAgger and progress-aware RL) is fed by a proprietary 4D+ motion capture pipeline that collects 200+ demonstrations per hour from live industrial environments, augmented ~5x via a synthetic data framework using video generation models. The model is benchmarked against NVIDIA GR00T N1.6 and Physical Intelligence π0.5, achieving SOTA on 8 simulation benchmarks and 86.8% on ALLEX real-world dexterous tasks, using ~20% of GR00T N1.5's training compute.
The company sells through a structured 'Robotics Transformation (RX)' go-to-market: an RX Diagnosis & Baseline assessment, a paid Proof-of-Concept in a controlled lab environment, and 3-to-5-year co-development partnerships that pair model licensing with continuous data capture. Revenue is generated via licensing/deployment of RLDX-1 across embodiments, professional services for engagement stages, and recurring subscription-style multi-year contracts. Named enterprise customers and pilot partners include CJ Logistics (logistics), Lotte Hotel Seoul (hospitality), Lawson via KDDI (retail), and Honda and BMW (manufacturing), while strategic investors include LG Electronics, SK Telecom, CJ Logistics, Lotte Ventures, Kakao Investment, Hanwha Asset Management, KDDI, ANA Holdings, and Mitsui Chemicals — investors that also provide operational training environments.
RLWRLD operates globally with headquarters in Seoul, regional headquarters in San Francisco led by President Carl Choi (ex-Alumni Ventures Deep Tech Fund), and an office in Tokyo. Strategic technology partnerships anchor its stack: NVIDIA (DexBench, Isaac Lab integration, full GR00T/Isaac Sim/Jetson/TensorRT stack), Microsoft (Azure cloud and go-to-market in Asia), AWS (Generative AI Accelerator cohort), and WIRobotics (ALLEX humanoid integration). The company has been recognized as a World Economic Forum 2026 Technology Pioneer and won first place in the Foundation Models category at the 2025 Nebius Robotics & Physical AI Awards, alongside major research acceptances at ICLR, CVPR, ICML, NeurIPS, and RSS.
RLWRLD firmographics
Firmographics- Name
- RLWRLD
- Legal name
- RLWRLD Inc.
- Website
- https://rlwrld.ai
- Company type
- Private
- Founded year
- 2024
- Operating status
- Operating
- Headcount range
- 51–100 employees
- Short description
- RLWRLD Inc. is a Seoul-based physical AI company building foundation models for dexterous robot manipulation, licensing RLDX-1 and running multi-year Robotics Transformation engagements with industrial enterprises in manufacturing, logistics, retail, hospitality, and airlines.
- Ownership category
- akta.pro rank
RLWRLD industry classification
Industry- Product category
- Robotics AI / Physical AI
- NAICS
- Custom Computer Programming Services (541511), Scientific Research and Development Services (5417), Research and Development in the Physical, Engineering, and Life Sciences (except Nanotechnology and Biotechnology) (541715)
- SIC
- Special Industry Machinery, Nec (3559)
- akta.pro primary industry
- Motion & Machine Control (CNC, Robotics Controllers, Servo/Motion Controllers) (IMAGABAG)
- akta.pro secondary industries
- Robotic Process Automation (RPA) & Digital Robotics (HDAAAIAI), Robotic Process Automation (RPA) & Hyperautomation Platforms (HDAEAKAF)
Keywords
Where RLWRLD is headquartered
LocationHeadquarters
- HQ city
- Seoul
- HQ country
- South Korea
- HQ region
- Asia
Offices3 records
Markets served
RLWRLD business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Marketing or Sales, Operations, Infrastructure
Revenue model
- Robotics Foundation Model Licensing & Deployment: RLWRLD develops and licenses its RLDX-1 robotics foundation model to enterprise customers in manufacturing, logistics, retail, hospitality, and airlines. Revenue is generated through long-term partnership agreements involving model fine-tuning for specific customer operations, co-building industry-specific models, and ongoing model improvement tied to accumulated operational data. The company operates as the 'intelligence layer' between robot hardware and end customers.
- Robotics Transformation (RX) Professional Services: Structured engagement model encompassing three stages: (1) RX Diagnosis & Baseline — operational task analysis with ROI measurement, (2) Proof-of-Concept validation in controlled lab environments, (3) Long-term industry model co-development partnerships spanning 3-5 years. Revenue generated through engagement fees at each stage with escalation to multi-year commitments.
- Data Partnership Model: Strategic partnerships with enterprises in logistics, manufacturing, and hospitality provide RLWRLD with access to real-world operational environments for training data collection, while partners benefit from robotic transformation. The model creates a compounding data flywheel: more data yields better models, which strengthens partnerships.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Hybrid | Multi-year contract | RX Diagnosis & Baseline — structured task analysis and ROI assessment |
| Hybrid | Pay-as-you-go | Proof-of-Concept (PoC) — focused task validation in controlled lab environment |
| Subscription | Multi-year contract | Long-Term Industry Model Partnership — co-development of fine-tuned industry model |
Go-to-market motion1 record
Distribution channels3 records
Marketing channels6 records
RLWRLD product offering
Product offeringCore offering
RLWRLD develops and licenses RLDX-1, an 8.1-billion-parameter dexterity-first robotics foundation model that enables humanoid and other robot embodiments to perform five-finger manipulation tasks for industrial use cases. The company collects proprietary 4D+ motion data from live factory, warehouse, retail, hospitality and airline environments, and delivers the model alongside a structured Robotics Transformation (RX) engagement methodology — diagnostics, proof-of-concept validation, and 3-to-5-year industry model partnerships.
Differentiator
Problem solved
Functional benefit
Brands
- All Hands Up!: A platform to analyze and visualize the performance of various dexterous robot hands using real-world operational data, utilizing the proprietary DexBench benchmark to evaluate robot hand characteristics across 18 practical operational tasks.
- DexBench
Products and services
- RLDX-1 Robotics Foundation Model 8.1-billion-parameter dexterity-first foundation model that gives humanoid and other robot embodiments five-finger manipulation capability for industrial tasks such as grasping, pouring, in-hand rotation, and tool use. Hardware-agnostic and deployable across multiple robot embodiments through fine-tuning. Sold to enterprise customers in manufacturing, logistics, retail, hospitality and airlines.
- DexBench Universal benchmark and data standard for evaluating humanoid robot dexterity performance across simulation and real-world settings. Co-developed with NVIDIA as an open industry specification and integrated with NVIDIA Isaac Lab and Isaac Lab-Arena. Used by robot manufacturers and research institutions to standardize dexterity measurement.
- All Hands Up! Platform Interactive platform that analyzes and visualizes dexterous robot hand performance using the DexBench benchmark and URDF-based interactive visualization. Enables manufacturers, researchers, and industry partners to compare robot hand characteristics across 18 practical operational tasks.
- Robotics Transformation (RX) Professional Services Three-stage professional services engagement for enterprise customers: (1) RX Diagnosis & Baseline — operational task analysis with ROI measurement; (2) Proof-of-Concept validation in controlled lab environments; (3) Long-term 3-5-year industry model co-development partnership with continuous data capture and model improvement.
Quantifiable outcome
- 86.8% average success rate on ALLEX functional capabilities vs 39.1% for Physical Intelligence π₀.₅ and 44.8% for NVIDIA GR00T N1.6
- +8 more outcomes
Companies that use RLWRLD
Customer profileNamed customers5 records
Segments6 records
Ideal customer profiles5 records
RLWRLD technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
Integration4 records
AI capability9 records
Feature10 records
RLWRLD partnerships and signals
Strategic signalPartnerships
Ten partnerships are on record, tiered core and significant.
- NVIDIAcoreStrategic collaboration to define next-generation standards for humanoid robot AI, focusing on three pillars: DexBench (universal dexterity benchmark), a common data standard for dexterous manipulation training, and deep integration with NVIDIA Isaac Lab and Isaac Lab-Arena frameworks. RLWRLD's RLDX-1 is built entirely on NVIDIA's robotics and AI stack (Isaac GR00T, Isaac Lab, Isaac Sim, cuRobo, H100/A100 GPUs, Jetson AGX Thor, TensorRT). NVIDIA CEO Jensen Huang publicly acknowledged RLWRLD as 'one of the core partners in the physical AI ecosystem we are building at NVIDIA.'
- Upstage (Sovereign AI Consortium)coreRLWRLD joined the Upstage-led consortium as part of South Korea's Sovereign AI Foundation Model initiative. Tasked with defining technical requirements for vision-language models in robot control applications and designing verification scenarios for industrial deployment. The consortium competes in government-led evaluation rounds to build globally competitive sovereign AI capabilities.
- LG ElectronicscoreStrategic investor providing capital and access to LG's manufacturing and electronics operations. LG Electronics announced at CES 2026 that it is adopting NVIDIA's GR00T and physical AI technologies, creating alignment with RLWRLD's NVIDIA-based foundation model approach.
- SK TelecomcoreStrategic investor providing capital and access to SK Telecom's technology ecosystem and infrastructure. Partner in South Korea's Sovereign AI Foundation Model initiative. Collaboration with SK Telecom, LG Electronics, and CJ Logistics to collect industrial data using RLWRLD's 4D+ capture technology.
- Mitsui ChemicalssignificantJapanese diversified chemicals company investing as a strategic partner. Provides access to manufacturing and chemical processing environments where dexterous manipulation is required, supporting RLWRLD's expansion into Japanese industrial sectors.
- MicrosoftcoreStrategic partnership to develop and commercialize industrial robotics AI technologies. RLWRLD leverages Microsoft Azure cloud infrastructure and Microsoft Research engagement for training, inference, and global deployment. Joint go-to-market efforts across Asia in manufacturing, logistics, retail, and hospitality sectors. The partnership combines RLWRLD's robotics foundation model trained on real-world data from Korea, Japan, and the US with Microsoft's global cloud reach.
- AWS (Amazon Web Services)coreRLWRLD selected for the third cohort of the AWS Generative AI Accelerator program, providing up to $1 million USD in AWS credits, personalized mentorship, and access to curated support network. AWS infrastructure supports RLWRLD's training and deployment workloads. CEO Junghee Ryu spoke at AWS re:Invent 2025 in Las Vegas about using physical AI robotics to address labor shortages.
- WIRoboticscorePartnership to integrate RLWRLD's intelligence into WIRobotics' ALLEX humanoid robot platform — the first general-purpose humanoid robot with 15-DoF five-finger hands (700g payload, 40N fingertip force) capable of human-like precision responses to physical stimuli. ALLEX serves as a primary mid-training embodiment for RLDX-1 (RLDX-1-MT-ALLEX checkpoint). RLWRLD's 'Dexterity Night' debut event featured ALLEX prominently.
- KDDIcoreJapanese telecommunications company (backed ANA Holdings and KDDI as co-investors) providing capital and access to Japanese convenience store chain Lawson for a proof-of-concept. KDDI has conducted a 3-month PoC with Lawson for shelf stocking robot AI validation. RLWRLD's Japan market strategy prioritizes logistics, service, and manufacturing sectors.
- ANA HoldingscoreJapan's largest airline operator invested via corporate venture capital alongside KDDI. Provides access to airline and aviation operational environments (cabin cleaning, galley support, passenger service) for robot training data and long-term deployment partnerships.
Scale indicators14 records
Recent moves8 records
Expansion highlights6 records
RLWRLD competitors and assessment
Company assessmentDirect peers
- Physical Intelligence: Directly comparable VLA/robotics foundation model developer (π0, π0.5). Both target dexterous manipulation as the core problem and benchmark head-to-head on RoboCasa and ALLEX; Physical Intelligence is RLWRLD's primary named performance benchmark competitor.
- Figure AI: Humanoid robot company building its own Helix foundation model for dexterous manipulation. Comparable in targeting the same dexterity-first commercial opportunity with vertically integrated hardware + AI, though Figure owns the robot fleet while RLWRLD is hardware-agnostic.
- 1X Technologies: Humanoid robot developer (Neo) with backing from OpenAI and EQT. Directly comparable in building AI-driven dexterous manipulation for humanoid platforms in real-world service and home settings; competing for the same enterprise customer mindshare.
- Skild AI: Robotics foundation model company building general-purpose brain for any robot embodiment. Closely comparable business model (hardware-agnostic foundation model + licensing) and at a similar stage of commercialization with strong VC backing.
- Agility Robotics: Digit humanoid robot commercializing in logistics (Amazon, GXO). Comparable in targeting the same dexterity/labor-replacement value proposition in warehouses and logistics, though Agility owns its fleet while RLWRLD licenses the intelligence layer.
- Covariant: AI foundation models for warehouse manipulation (acquired by Amazon in 2024). Closely comparable in providing AI-powered dexterous picking solutions for industrial logistics; represents the acquired/incumbent path that RLWRLD's category leaders have taken.
Broad incumbents
- NVIDIA (GR00T / Isaac): Both partner and competitor. NVIDIA GR00T N1.6 is the named benchmark target for RLDX-1; Isaac Lab is the shared simulation platform. NVIDIA controls the underlying compute and simulation stack RLWRLD is built on, creating a vertically-integrating incumbent dynamic.
- Boston Dynamics: Established humanoid (Atlas) and mobile manipulation robot developer. Comparable end-customer overlap in industrial and logistics automation, but Boston Dynamics is vertically integrated with its own hardware and Atlas AI stack rather than licensing a standalone foundation model.
Emerging players
- Toyota Research Institute (TRI): Large-scale behavior robotics research with LBM (Large Behavior Model) for dexterous manipulation. Comparable in publishing SOTA dexterous manipulation results, but operates as a research arm rather than a commercial foundation model vendor.
- WIRobotics: Korean humanoid robot company (ALLEX) and RLWRLD's primary mid-training embodiment partner. ALLEX is one of the few commercial 15-DoF 5-finger humanoid platforms; comparable as a hardware-native Korean robotics peer even though the relationship is partnership-driven.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat4 records
Key risks7 records
Key highlights7 records
Customer concentration
RLWRLD social profiles
Digital presenceRLWRLD financial estimates
Financial estimateRevenue estimate
Valuation estimate
RLWRLD leadership team
Management profileNumber of profiles
Profiles4 records
RLWRLD funding detail
Funding detailFunding overview
Funding rounds3 records
Investors23 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
RLWRLD M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about RLWRLD
What does RLWRLD do?
RLWRLD develops and licenses RLDX-1, an 8.1-billion-parameter dexterity-first robotics foundation model that enables humanoid and other robot embodiments to perform five-finger manipulation tasks for industrial use cases. The company collects proprietary 4D+ motion data from live factory, warehouse, retail, hospitality and airline environments, and delivers the model alongside a structured Robotics Transformation (RX) engagement methodology — diagnostics, proof-of-concept validation, and 3-to-5-year industry model partnerships.
Is RLWRLD a public or private company?
RLWRLD is a private company. It is classified as venture growth investor backed and is currently operating.
When was RLWRLD founded?
RLWRLD was founded in 2024. It employs 51 to 100 people.
Where is RLWRLD based?
RLWRLD is headquartered in Seoul, South Korea, in the Asia region.
How does RLWRLD make money?
Three revenue lines are on record. Robotics Foundation Model Licensing & Deployment is the primary driver. The others are robotics Transformation (RX) Professional Services and data Partnership Model.
Who are RLWRLD's main competitors?
Direct peers on record are Physical Intelligence, Figure AI, 1X Technologies, Skild AI, Agility Robotics and Covariant. Broad incumbents are NVIDIA (GR00T / Isaac) and Boston Dynamics. Emerging players are Toyota Research Institute (TRI) and WIRobotics.
Does RLWRLD have an API?
No public API is recorded for RLWRLD.
What industry is RLWRLD in?
RLWRLD's product category is Robotics AI / Physical AI. Its primary akta.pro industry code is IMAGABAG, Motion & Machine Control (CNC, Robotics Controllers, Servo/Motion Controllers), with a secondary code of HDAAAIAI, Robotic Process Automation (RPA) & Digital Robotics. Its NAICS code is 541511 and its SIC code is 3559.