Volumes

What Volumes does

Volumes is an early-stage startup headquartered in New York City that captures ground-truth volumetric data from real-world environments and licenses it to AI labs for training physical AI and robotics systems. The company operates proprietary multi-sensor fusion rigs that record synchronized spatial geometry, temporal dynamics, environmental context, and physical interactions (force and contact) with microsecond precision, then processes these captures into AI-ready formats.

The product portfolio spans a four-stage data pipeline: (1) Raw Ground-Truth Volumetric Data — uncompressed multi-sensor captures delivered via cloud streaming or physical drives; (2) 3D Gaussian Splatting Data — high-fidelity static 3D reconstructions rendering in real time at 100+ FPS; (3) 4D Gaussian Splatting Data — spatiotemporal reconstructions encoding geometry, photorealistic appearance, and temporal dynamics for simulation, rendering, and AI training, with multiple export formats including PLY and CUDA-ready; and (4) Synthetic Volumetric Data — AI-generated scenes with controllable variations in lighting, weather, and environment, claimed to be 10x more cost-effective than real-world capture while preserving physical accuracy.

Volumes runs a sales-led B2B motion targeting AI labs and robotics/autonomous systems companies, with quote-based enterprise pricing and no self-serve or freemium option. The company was founded in 2026 by Chet Ellis (CEO), Tyler Raciti (COO), and a CTO who remains in stealth pending public announcement. The team is small, the company is actively recruiting across computer vision, robotics, and volumetric data, and no funding rounds, named customers, or revenue figures have been disclosed.

Name

Volumes

Legal name

Volumes

Website

https://volumes.cloud

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

1–10 employees

Short description

Volumes is an early-stage NYC startup that captures ground-truth volumetric data from real-world scenes using proprietary multi-sensor rigs and 4D Gaussian splatting, licensing it to AI labs for physical AI and robotics training.

Ownership category

akta.pro rank

Product category

AI Training Data Infrastructure

NAICS

Postproduction Services and Other Motion Picture and Video Industries (51219)

SIC

Services-Computer Processing & Data Preparation (7374)

akta.pro primary industry

Data Pipelines for GenAI (Curation, Filtering, Deduplication, Copyright) (HDAAACAJ)

akta.pro secondary industries

3D Vision, Depth Sensing & Point Cloud Analytics (HDAAAEAK), Real-Time Environment/Asset Creation & Photogrammetry (MPAHADAI), Photogrammetry & Reality Capture for AM (IMAAAHAB)

Keywords

• Volumetric data capture
• Physical AI training
• 4D Gaussian splatting
• Synthetic data generation
• Multi-sensor fusion

Headquarters

HQ city

New York

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Operations, Infrastructure

Revenue model

1. Volumetric Data Licensing: Licensing of ground-truth volumetric data products including raw captures, 3D/4D Gaussian splatting data, and synthetic volumetric scenes to AI labs for training physical AI systems. Quote-based pricing with flexible delivery options including cloud streaming or physical drives.

Go-to-market motion1 record

Distribution channels1 record

Marketing channels3 records

Core offering

Volumes captures ground-truth volumetric data from real-world scenes using proprietary multi-sensor fusion rigs and processes it into formats suitable for AI training. Their product pipeline progresses from raw multi-sensor captures through static 3D Gaussian splatting reconstructions, spatiotemporal 4D Gaussian splatting representations, and AI-generated synthetic volumetric data. The data is licensed to AI labs and physical AI companies to train robotics and autonomous systems.

Product overview

Volumes offers a unified data platform for physical AI with four distinct products forming a data pipeline from capture to AI-ready training sets: Raw Ground-Truth Volumetric Data (unprocessed multi-sensor captures from proprietary sensor fusion rigs), 3D Gaussian Splatting Data (high-fidelity static 3D scene reconstructions), 4D Gaussian Splatting Data (spatiotemporal reconstructions capturing space and time), and Synthetic Volumetric Data (AI-generated scenes expanding training datasets). The products progress from raw capture through reconstruction to AI-ready synthetic data, enabling AI labs to license accurate, licensable ground-truth data for training physical AI systems.

Differentiator

Problem solved

Functional benefit

Products and services

• Raw Ground-Truth Volumetric Data

Quantifiable outcome

• 10x cost reduction in training data generation compared to real-world capture

Segments2 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability6 records

Feature4 records

Scale indicators1 record

Recent moves5 records

Expansion highlights4 records

Direct peers

• World Labs: World Labs (Fei-Fei Li) is building 'spatial intelligence' foundation models from 3D/4D world data. It is the most direct competitor in marrying Gaussian-splatting-class capture with foundation-model ambitions for physical AI.
• Encord: Encord provides data infrastructure for curating, labelling, and managing AI training datasets (vision and multimodal). Volumes' data-pipeline product competes for the same AI-data workflow budget.
• Luma AI: Luma AI builds 3D capture and Gaussian splatting technology (Genie, Dream Machine) for generating 3D/4D content from video and images. Volumes overlaps directly on Gaussian splatting-based scene reconstruction and the multi-modal capture pipeline targeting AI training.
• Polycam: Polycam is a photogrammetry and Gaussian splatting platform for capturing 3D scenes from phone and LiDAR. It is the closest direct competitor on the volumetric-capture tooling layer.
• Synthesis AI: Synthesis AI generates synthetic visual data (faces, objects, environments) for training computer vision models. Volumes is the 3D/4D volumetric analogue, selling synthetic volumetric scenes to physical-AI labs.
• Scale AI: Scale AI is the dominant data infrastructure and labelling provider for AI training. Volumes sells a similar 'training data as a product' value proposition to the same AI-lab buyer persona, albeit specialised in physical/3D/4D ground-truth rather than text/image labels.

Emerging players

• Skild AI: Skild AI builds foundation models for robotics and physical AI — an adjacent buyer of Volumes' data and a potential customer or competitor for the downstream model layer.
• Niantic: Niantic operates a large-scale 3D scanning/AR mapping stack from Pokémon Go and Scaniverse. It represents an alternative path to 3D-world capture infrastructure for training AI.
• Physical Intelligence (PI): Physical Intelligence builds generalist robot foundation models and is one of the highest-spend buyers of physical-world training data, making it a logical target customer for Volumes.
• Kognic: Kognic specialises in annotated and curated perception datasets for autonomous driving and robotics — a vertical overlap with Volumes' robotics/autonomous-systems customer segment.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat3 records

Key risks7 records

Key highlights7 records

Customer concentration

• LinkedInlinkedin.com/company/volumes-ai

Revenue estimate

Valuation estimate

Number of profiles

Profiles3 records

Funding overview

Funding rounds1 record

Investors1 record

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Volumes