01
Ego/Exo4D Datasets
Egocentric and exocentric 4D capture, performed by trained, fairly-paid operators. High-fidelity fuel for robotics, physical AI, and frontier labs.
Learn moreAbout us
Human skill, refracted into the data that powers what's next.
Diffraction is a vertically integrated AI training-data company built on a simple belief: people are the most valuable dataset on earth. From our home base in Columbus, Ohio to communities around the world, we manufacture certified, instrumented training data — we don't scrape it. Real human skill becomes the data behind robotics, physical AI, and the next generation of creator economies. We empower the workforce and the data follows.
2022
Founded
500+
Trained humans
$700k+
Paid out
150k+
Hours in pipeline
Five pillars, one mission
02
ACT Policy Training
Demonstration data for advanced manipulation and navigation policies, captured in real-world conditions and tuned for deployment. LeRobot and GR00T policy exports available.
Learn more03
StreamGenie.ai
A new income layer for creators: license your content to frontier AI models through a marketplace built for specialized requests, while keeping full ownership of everything you make.
04
Influencer Marketing
A vetted roster and end-to-end campaign management. We pair brands with creators who actually move the needle, then run the entire campaign from first brief to final report.
05
Creator Network
300+ TikTok LIVE creators backed by coaching, tools, and a community that shows up for each other. Free to join and we help you grow your LIVE audience.
Manufactured, not found.
Everyone else sells found data. We manufacture certified, instrumented training data — captured by consenting, fairly-paid people and processed on one factory line.
A signed birth certificate
Every dataset ships with a cryptographically signed certificate proving consent, licensing, and provenance — revocation-aware, with EU AI Act transparency baked in. Scrapers structurally cannot retrofit consent onto data they took.
Captured to spec
Purpose-built pod studios and a 300+ consented creator network capture what buyers actually ask for — not whatever happened to exist. Spec in, certified dataset out.
One pipeline, two products
The same five-step factory line — ingest, enrich, certify, QA, export — serves licensed creator video at volume and heavy, instrumented pod capture for robotics labs.
Holding co.
Diffraction
The operating umbrella: creator development frameworks, partnerships, and the legal and contractual layer that makes the rest possible.
Supply moat · est. 2022
TikTok LIVE Creator Network
An official US partnership program serving 300+ creators with free coaching, analytics, and monetization. We built the relationships before we built the product.
Growth engine · launched 2025
StreamGenie.ai
The data platform: ingestion, annotation, a live two-sided marketplace, and pod-room infrastructure.
Our team
Humans are at the center of everything we do. Founded in Columbus, Ohio in 2022, we believe the path to better AI runs straight through people. Their skills, their stories, their communities. So that's what we build: real opportunity for workers and creators, here in Ohio and around the world.
Thom Swann
Co-Founder
Writing production code since 1999. Built StreamGenie's full stack — ingestion, segmentation, annotation, QA, and marketplace — after a decade of live-streaming tools used by thousands of streamers and leading Wendy's digital ordering.
Joseph Sottile
Co-Founder
Creator-economy operator who's worked every side of the table — streamer, talent manager, agency founder. Has managed creators like Linus Tech Tips and FaZe Rug; runs our TikTok LIVE network and Diffraction Academy.
Claire
Operator
Trained capture operator specializing in Ego/Exo4D datasets.
Humans are at the center of everything we do.
Every dataset we capture starts with a person mastering a skill. Every campaign we run starts with a community we've earned. Real people, real relationships, real results.