Config
Config builds data infrastructure and a proprietary foundation model (CFG-1) for general-purpose bimanual robotic manipulation, serving enterprise manufacturers via the Loop platform and Catalog dataset product, with operations in the US, South Korea, and Vietnam.
- Company typePrivate
- Founded2025
- HeadquartersSan Jose, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What Config does
Config builds the data infrastructure and foundation models that enable general-purpose bimanual robotic manipulation. The company's core offering is the Loop platform, which turns real-world human and robot demonstrations into deployable, task-specific robot policies through six composable modules (collect, curate, convert, train, deploy, evaluate) backed by two proprietary GPU engines — the Conversion Engine, which translates human action data into robot-ready trajectories at micrometer precision, and the Policy Engine, which hosts Config's own CFG-1 foundation model (a 2B-parameter autoregressive transformer with 3-minute native memory and sub-50ms inference on a single RTX 5090) alongside open-source models such as Pi 0.5 and NVIDIA GR00T N1.6. Underpinning Loop is an in-house data corpus of 200,000+ hours across 10,000+ tasks, generated through a Seoul teleop lab and a Hanoi Data Studio employing approximately 300 local workers producing roughly 20,000 hours of human action data per month; the same corpus is sold separately through the Catalog dataset product in LeRobot format.
Config monetizes through enterprise sales of the Loop platform to large manufacturers and Fortune Global 500 automation buyers (quote-based, multi-year engagements scoped by Forward-Deployed Engineers), one-time dataset sales via Catalog, and billable custom data collection and model adaptation services; a cloud-based robot-as-a-service product is on the stated roadmap targeting $10M ARR by 2027. The company is headquartered in San Jose with a Seoul engineering HQ and operates its Hanoi data factory under the legal entities Config Intelligence Inc. (US) and Config Intelligence Korea (KR), having closed a $27M oversubscribed seed round in May 2026 at a valuation exceeding $200M, led by Samsung Venture Investment with Hyundai's ZER01NE Ventures, LG Technology Ventures, and SKT America.
Config firmographics
Firmographics- Name
- Config
- Legal name
- Config Intelligence Inc.
- Website
- https://config.inc
- Company type
- Private
- Founded year
- 2025
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- Config builds data infrastructure and a proprietary foundation model (CFG-1) for general-purpose bimanual robotic manipulation, serving enterprise manufacturers via the Loop platform and Catalog dataset product, with operations in the US, South Korea, and Vietnam.
- Ownership category
- akta.pro rank
Config industry classification
Industry- Product category
- Robotics Data Infrastructure
- NAICS
- Computer Systems Design and Related Services (54151), Custom Computer Programming Services (541511), Computer Systems Design Services (541512)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Integrated Systems Design (7373)
- akta.pro primary industry
- Robot Software, Autonomy Stacks & Simulation (ROS, digital twins) (HDAAAIAK)
- akta.pro secondary industries
- Humanoid & General-Purpose Robots (HDAAAIAJ), Robotics Kits, Parts & Lab Equipment Providers (Education-Focused) (EDAMADAN)
Keywords
Where Config is headquartered
LocationHeadquarters
- HQ city
- San Jose
- HQ country
- United States
- HQ region
- North America
Offices3 records
Markets served
Config business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Infrastructure, Operations, Marketing or Sales
Revenue model
- Enterprise Platform Sales (Config Loop): Direct B2B sales of the Config Loop platform — six composable data-and-model modules delivered to enterprise manufacturers and Fortune Global 500 customers. Engaged via enterprise sales cycle with Forward-Deployed Engineers scoping PoCs, pilots, and production deployments.
- Data Licensing (Catalog): Catalog — ready-made, pre-converted human action and robot teleop datasets (200K+ hours, 10K+ tasks, 10M+ episodes) sold outright as LeRobot-format datasets; usable for pre-training and post-training without operating any pipeline.
- Cloud-based Robot-as-a-Service (planned): Roadmap item from the seed funding announcement: launch a cloud-based robot-as-a-service product, scaling toward 1 million hours of collected motion data and $10 million ARR by 2027.
- Custom Data Collection & Model Adaptation Services: Forward-Deployed Engineers scope and deliver task-specific data collection, model fine-tuning/adaptation, and evaluation as part of customer engagements — billable professional/managed services revenue tied to deployment cycles.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Other | Multi-year contract | Enterprise sales — quote-based for Config Loop deployments |
| One time/ perpetual license | — | Catalog dataset purchases — pre-converted human and robot teleop data |
| Usage-based | Pay-as-you-go | Planned cloud-based Robot-as-a-Service |
Go-to-market motion5 records
Distribution channels5 records
Marketing channels7 records
Config product offering
Product offeringCore offering
Config builds the data infrastructure and technology for general-purpose bimanual robotics. Its core product, the Loop platform, turns real-world demonstrations into deployable, task-specific robot policies through six composable modules (collect, curate, convert, train, deploy, evaluate) and two GPU engines (Conversion Engine and Policy Engine). Config also sells CFG-1, a 2B-parameter robot foundation model for bimanual manipulation, and Catalog, a buy-outright dataset product offering pre-converted human and robot teleop data in LeRobot format.
Product overview
Config is a single-product company organized as a platform-plus-modules architecture anchored by three named offerings: (1) Loop, the core data-infrastructure platform with six composable modules (collect, curate, convert, train, deploy, evaluate) backed by two GPU engines (the Conversion Engine for human-to-robot trajectory conversion and the Policy Engine for training and live inference); (2) CFG-1, Config's first-generation 2B-parameter robot foundation model for bimanual manipulation that is hosted inside Loop's Policy Engine and powers every deployment; and (3) Catalog, a separate buy-outright dataset product offering 200K+ hours of pre-converted human and robot data drawn from Config's own pipeline. Customers engage with one or more products depending on maturity: those who want ready-made data use Catalog, while those building policies run the full Loop pipeline — with CFG-1 or open-source models like Pi 0.5 or GR00T N1.6 as the underlying policy model.
Differentiator
Problem solved
Functional benefit
Brands
- Loop: Config's end-to-end platform that turns real-world demonstrations into deployable, task-specific robot policies across six composable modules (collect, curate, convert, train, deploy, evaluate) and two GPU engines.
- Catalog
- CFG-1
Products and services
- Loop Config's core data-infrastructure platform that turns real-world demonstrations into deployable, task-specific robot policies. Built around six composable modules (collect, curate, convert, train, deploy, evaluate) and two GPU engines (Conversion Engine and Policy Engine), with four engagement models: full pipeline, collect+curate, collect+curate+convert, or Catalog-only data. Targets enterprise manufacturers and automation buyers.
- CFG-1 Config's first-generation robot foundation model for bimanual manipulation — a 2B-parameter autoregressive transformer with a linear action head, 3-minute native memory, and sub-50 ms inference on a single RTX 5090. Trained on a mix of human demonstrations (converted via the Conversion Engine) and high-precision robot teleop data; reaches 80%+ task success with just 1 hour of task-specific teleop fine-tuning. Powers every Config Loop deployment.
- Catalog A separate, buy-outright dataset product drawing from Config's pipeline. Offers 200K+ hours of pre-converted human action data (Hanoi Data Studio) for pre-training, paired with high-precision robot teleop data (Seoul lab) for post-training, across 10K+ tasks and 10M+ episodes. Delivered in LeRobot format, browsable by source, embodiment, and task.
Companies that use Config
Customer profileSegments3 records
Ideal customer profiles2 records
Config technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration7 records
AI capability10 records
Feature7 records
Config partnerships and signals
Strategic signalPartnerships
Five partnerships are on record, tiered flagship and core.
- LG CNSflagshipLG CNS launched PhysicalWorks, an end-to-end robot transformation (RX) platform, and named Config as a strategic partner for industry-specific robot foundation models (RFMs) inside its RX Innovation Lab. Config is one of LG CNS's strategic US robotics investments (alongside Dexmate and Skilled AI); Config's foundation models are positioned inside LG CNS's enterprise robot deployments across Korea.
- MassRoboticscoreMassRobotics, AWS, and NVIDIA jointly run the Physical AI Fellowship — Config was selected to the second cohort (March 12, 2026). Fellows receive engineering support, cloud infrastructure, AWS credits, and access to MassRobotics' Boston facilities, with the cohort showcased at the Robotics Summit & Expo Boston on May 27-28, 2026.
- Amazon Web Services (AWS)flagshipConfig's data pipeline runs natively on AWS — processing more than 100TB of robot data per month with 6x lower transfer cost and 2x lower storage cost vs Config's prior setup. Config's Loop is architected on AWS EKS, Terraform IaC, and Hyperpod-ready infrastructure; Config is a co-sponsor (via AWS Startups) of the MassRobotics Physical AI Fellowship and a featured startup in AWS Korea's 'Physical AI Zone' and 'Physical AI Frontier' program at AWS Summit Seoul 2026.
- NVIDIAcoreCo-sponsor (via NVIDIA Inception) of the MassRobotics Physical AI Fellowship; Config fellows gain access to NVIDIA's robotics tools and frameworks including the Isaac platform and Cosmos foundation models. RTX 5090-class GPUs are Config's inference target for CFG-1 deployment.
- Samsung C Lab OutsidecoreConfig was selected for the 8th cohort of Samsung's C Lab Outside, an open-innovation program that pairs startups with Samsung for business meetings and proof-of-concept trials in real industrial settings. The relationship gave Config hands-on experience producing data for actual industrial sites.
Scale indicators14 records
Recent moves7 records
Expansion highlights6 records
Config competitors and assessment
Company assessmentOthers
- Hugging Face (LeRobot): Open-source ecosystem provider whose LeRobot format is the standard Config uses to deliver Catalog datasets. Comparable as a thematically adjacent enabler in the open-source robotics stack that Config benefits from but does not directly compete with.
Broad incumbents
- Figure AI: San Jose-based humanoid robot developer with its own Helix foundation model for manipulation. Comparable as a broader incumbent that has built a vertically integrated manipulation model stack similar in spirit to CFG-1.
- NVIDIA (GR00T / Isaac): Provider of GR00T N1.6 — a humanoid robot foundation model hosted alongside CFG-1 inside Config's own Policy Engine — plus Isaac Sim and Cosmos simulation frameworks. Comparable as a broad incumbent in robot foundation models whose ecosystem Config plugs into rather than competes with head-on.
- Boston Dynamics: Waltham-based humanoid (Atlas) and quadruped robotics company with the Reinforcement Learning Research Kit and ongoing work in manipulation models. Comparable as a broad incumbent with proprietary manipulation IP that competes for the same enterprise automation budgets.
Emerging players
- Sanctuary AI: Vancouver-based humanoid robotics company developing Carbon, a large behavior model for dexterous manipulation. Comparable as an emerging player building foundation-model-driven manipulation policies on top of in-house hardware.
- Toyota Research Institute (Large Behavior Models): Research institute that originated diffusion-policy and large behavior model approaches for robot manipulation. Comparable as an emerging player whose published methods directly influence the architecture Config and others build on.
Direct peers
- Physical Intelligence: San Francisco-based developer of foundation models for general-purpose robot manipulation (π0). Most directly comparable peer to Config — both build general-purpose manipulation policies and have raised at multi-hundred-million-dollar valuations on similar narratives.
- Skild AI: Pittsburgh-based 'robot brain' startup building foundation models for embodied AI, co-invested in by LG CNS alongside Config. Comparable in building general-purpose manipulation and mobility intelligence that OEMs and SIs can integrate into their hardware.
- Covariant: Emeryville-based developer of AI foundation models for robotic manipulation in warehouse and industrial settings. Comparable in building manipulation policies that adapt to new SKUs and tasks using imitation and foundation-model techniques.
- Dexterity: Redwood City-based provider of AI-powered robotic manipulation systems for warehousing, parcel, and supply-chain applications. Comparable as a manipulation-focused AI company targeting industrial B2B deployments with task-specific robot policies.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat7 records
Key risks7 records
Key highlights6 records
Customer concentration
Config social profiles
Digital presenceConfig compliance and trust
Trust signalCompliance3 records
Config financial estimates
Financial estimateRevenue estimate
Valuation estimate
Config leadership team
Management profileNumber of profiles
Profiles2 records
Config funding detail
Funding detailFunding overview
Funding rounds1 record
Investors9 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Config 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 Config
What does Config do?
Config builds the data infrastructure and technology for general-purpose bimanual robotics. Its core product, the Loop platform, turns real-world demonstrations into deployable, task-specific robot policies through six composable modules (collect, curate, convert, train, deploy, evaluate) and two GPU engines (Conversion Engine and Policy Engine). Config also sells CFG-1, a 2B-parameter robot foundation model for bimanual manipulation, and Catalog, a buy-outright dataset product offering pre-converted human and robot teleop data in LeRobot format.
Is Config a public or private company?
Config is a private company. It is classified as venture growth investor backed and is currently operating.
When was Config founded?
Config was founded in 2025. It employs 1 to 10 people.
Where is Config based?
Config is headquartered in San Jose, United States, in the North America region.
How does Config make money?
Four revenue lines are on record. Enterprise Platform Sales (Config Loop) is the primary driver. The others are data Licensing (Catalog), cloud-based Robot-as-a-Service (planned) and custom Data Collection & Model Adaptation Services.
Who are Config's main competitors?
Hugging Face (LeRobot) is listed as an others. Broad incumbents are Figure AI, NVIDIA (GR00T / Isaac) and Boston Dynamics. Emerging players are Sanctuary AI and Toyota Research Institute (Large Behavior Models). Direct peers are Physical Intelligence, Skild AI, Covariant and Dexterity.
Does Config have an API?
Yes. The Config Loop client reaches the GPU engines (Conversion Engine and Policy Engine) over an API. The Policy Engine "serves them on your robot, running locally on an RTX 5090-class GPU or on a remote server, reached by the Loop client over an API." The Loop client is the same whether an engine runs on the user's bench or on a remote server. No public developer documentation URL, SDK download, or rate-limit information is provided; the API appears to be the platform's own client-to-engine protocol rather than a public open developer API.
What industry is Config in?
Config's product category is Robotics Data Infrastructure. Its primary akta.pro industry code is HDAAAIAK, Robot Software, Autonomy Stacks & Simulation (ROS, digital twins), with a secondary code of HDAAAIAJ, Humanoid & General-Purpose Robots. Its NAICS code is 54151 and its SIC code is 7372.