Poseidon Raises $15M Seed Round to Accelerate Physical AI
July 22, 2025
Founded by Sandeep Chinchali and Sarick Shah, incubated by Story, and backed by $15M from a16z crypto, Poseidon is building the infrastructure to collect, curate, and license the real-world data needed to accelerate physical AI innovation.
The Compute Race is Over. The Model Wars Are Ending. The Next Frontier is Data.
The first wave of generative AI was trained on the open web — books, Reddit threads, Wikipedia pages, and more. While those sources helped catalyze early breakthroughs, they’re now largely depleted.
The legal landscape is still evolving: some recent rulings have supported fair use in AI training, but the uncertainty around copyright, provenance, and commercial licensing remains a real concern for those building enterprise systems.
As AI advances, it has become clear to the world’s leading AI companies that scraped data isn’t designed for physical AI. Synthetic data can’t simulate edge cases, synthesize cultural nuance, or assume all the detail from the physical world that is needed to train robots, autonomous vehicles, voice systems, and more.
The next generation of AI systems will rely on something else entirely.
The Physical World Is the New Dataset
We’re moving from language-only models to systems that need to interpret environments, emotions, and actions. That shift demands high volumes of one of the most valuable subsets of IP: data from the real world. This data needs to be collected, curated, and licensed.
Some examples include:
First-person POV video of household chores to train robotic systems.
Multilingual speech data across varied accents and intonations to improve speech-to-text.
Sensor-rich driving footage from rare edge-case scenarios to inform self-driving cars.
The data exists, but it’s fragmented, living on phones, dashcams, GoPros, homes, and warehouses. No one has aligned incentives to coordinate the collection and licensing of this data at scale, until now – made possible by Story’s IP infrastructure.
Poseidon Is the Infrastructure Layer for AI Data
Poseidon is building the rails to connect AI teams with the exact training data they need, ensuring that they are ready for commercial use.
Poseidon is built on four core beliefs:
Demand-first design: Start with what AI companies actually need, not what contributors might happen to provide.
Decentralized scale: Real-world diversity requires distributed data creation. Poseidon makes it scalable.
Structured and validated: All data is curated, cleaned, labeled, and enriched for model training pipelines.
IP licensed by default: Every asset is registered via Story, with full traceability and licensing.
Built by Veterans, Backed by the Best
Built on and incubated by Story, Poseidon is proud to announce a $15M seed round led by a16z crypto to coordinate the aggregation of one of the most valuable forms of IP – high-quality data – to accelerate AI innovation.
"AI foundation models have already exhausted the most easily accessible training data. Poseidon's decentralized data layer seeks to establish a new economic foundation for the internet, rewarding creators and suppliers for providing the diverse inputs that next-gen intelligent systems need. We are excited to support Poseidon in its work to solve one of the most critical bottlenecks in AI development." – Chris Dixon, Founder and Managing Partner, a16z crypto
Stewarded by Story Founder and CEO, Seung Yoon “S.Y.” Lee as the project’s founding President, Poseidon co-founders Sandeep Chinchali and Sarick Shah bring deep, cross-disciplinary expertise in AI, systems, and applied science.
Sandeep is a Stanford PhD, AI researcher and assistant professor at UT Austin, where he leads a lab focused on edge computing, networked robotics, and generative AI. With 1,500+ citations, Sandeep’s work bridges theoretical research and real-world systems, helping make AI more practical and performant across physical environments. As Chief AI Officer at Story, Sandeep helped design the IP infrastructure that now underpins Poseidon’s licensing and provenance systems.
Sarick is a product-driven AI engineer and has built and deployed AI systems across telecom, finance, and logistics—from causal inference and time-to-event RNNs at LotusFlare to multi-agent fleet management systems at Roadz. Most recently, Sarick was the Lead AI Engineer at Story, driving several initiatives including influence function research and the adoption of natural language interfaces (eg. MCP servers) to seamlessly register and search IP on the Story blockchain.
Seung Yoon “SY” Lee served as Global Strategy Officer at Kakao Entertainment, a leading Korean entertainment company, after selling his first venture, Radish – a mobile serialized fiction app with millions of downloads – for $440M in 2021. A serial entrepreneur and venture partner at Hashed, he has been recognized as an Asia Society’s Asia 21 Young Leader, inaugural member of Forbes 30 under 30 Asia, a Trilateral Commission David Rockefeller Fellow, and the first Asian President of the Oxford Union.
A Full-Stack Solution for a Data-Constrained World
Poseidon’s architecture covers the entire lifecycle of AI data and is designed for enterprise-grade usage:
Collection: From smartphone SDKs to specialized DePIN apps, Poseidon makes distributed collection easy.
Curation: Poseidon’s ML pipelines handle format standardization, PII removal, duplication checks, and quality scoring.
Labeling: Foundation model-powered labeling pipelines route edge cases to human reviewers only when needed.
IP Registration: Every dataset is registered as an IP asset on Story for traceability and enforcement.
IP Management
Ensuring that datasets of this scale are fully licensed and compliant are only possible with Story’s IP infrastructure. Every data point entering the Poseidon network is registered as an IP asset on Story's blockchain, creating an immutable record of its source, licensing terms, and chain of custody. This addresses the IP safety concerns that increasingly dominate enterprise AI procurement.
Robotics Is Just the Start
Poseidon's initial focus will be on curating training data for robotics, specifically egocentric, POV data. Why robotics? Because their needs are urgent, their data is cross-applicable (CV, video gen, 3D modeling), and the demand has been confirmed by leading robotics teams.
In addition to robotics, research and development is underway for the addition of audio, biometric, and healthcare data to Poseidon’s data stack.
Poseidon gives teams the foundation they need to build systems that will change the world.
Model architecture is open-source. Compute is a commodity. Data is the new moat.
Welcome to the future of data infrastructure for AI.