Andrew Dai's stealth venture targets the visual reasoning gap in AI, backed by NVIDIA, Menlo Ventures, and Google's Jeff Dean
Andrew Dai spent 14 years at Google DeepMind helping build some of the foundational research that eventually informed systems like ChatGPT. Now he’s betting that the entire AI industry has been looking at intelligence through the wrong lens, literally.
His new startup, Elorian AI, just emerged from stealth with $55 million in seed funding at a $300 million post-money valuation. The round was co-led by Striker Ventures and Menlo Ventures, with Altimeter Capital, NVIDIA, and former Google Senior Fellow Jeff Dean also participating.
What Elorian actually does #
Here’s the core thesis: most AI systems today, even the ones that can “see,” process visual information by first translating it into text. Think of it like someone describing a painting to you over the phone versus actually looking at it yourself. You lose spatial relationships, physical context, and the kind of intuitive reasoning that humans do effortlessly when they look at the world.
Elorian wants to build AI that reasons natively in the visual medium. No text intermediary. No lossy translation step.
The company calls this the “reasoning gap” in current AI models. Most systems treat visual information as static data, a snapshot to be labeled and categorized. Elorian’s approach targets something more ambitious: dynamic visual reasoning that enables long-horizon planning and causal inference in real-world environments.
Co-founder Yinfei Yang, who previously worked on multimodal systems at Apple, rounds out the technical leadership. Together, Dai and Yang are targeting applications across robotics, autonomous agents, industrial inspection, aerospace, and medicine.
The crypto connection (or lack thereof) #
Elorian has no blockchain component, no token, and no decentralized anything in its tech stack. There is an unaffiliated meme token called ELORIAN trading on the Solana blockchain. It has nothing to do with the company. This is worth mentioning only because it’s the kind of thing that catches retail traders off guard. Someone sees a headline about a $300 million AI startup, finds a token with the same name, and mistakes correlation for connection. It’s not.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our