Former Google DeepMind Researcher Raised $55 Million Before Building a Product

Andrew Dai spent more than a decade at Google Brain and Google DeepMind, co-authoring research that later influenced the development of ChatGPT. Just months after leaving the company, with no finished product in hand, he raised $55 million for his new startup, Elorian, at a $300 million valuation. TechCrunch reported the story behind the round on Wednesday.

The gap in visual understanding

Elorian's core thesis is simple: today's best AI models, including ones Dai himself helped build, excel at math, coding, and text generation, but stumble on basic visual tasks. They can't reliably count objects in a photo, match shapes, or reason about what they're actually looking at.

You have models that are really, really good at math, really good at coming up with new ideas in physics, and of course coding is very popular right now... But one of the areas where progress has been remarkably uneven is visual understanding and visual reasoning - Andrew Dai, founder and CEO of Elorian

Dai calls this the visual AGI gap. The idea is that language models have learned to represent the world mainly through text and code, while the visual layer, actually perceiving and reasoning about images, has fallen behind. Elorian is meant to operate as a research-and-product lab focused exclusively on multimodal reasoning, with a particular emphasis on vision.

Who is behind the company

The company was founded in 2025. Besides Dai, its co-founder is Yinfei Yang, formerly chief scientist on Apple's machine learning team. Both have careers spent at companies that defined the last decade of AI development, which for investors was reason enough to back them without seeing a working prototype.

The seed round was led by Menlo Ventures and Altimeter, with additional backing from 49 Palms. Among the individual investors was Jeff Dean, one of Google's most influential engineers, itself a signal of the research community's confidence in the Elorian team.

Money before product

A $300 million valuation with no finished product shows just how much venture capital is willing to pay today for a research team's reputation alone, provided it comes out of top AI labs. It's a pattern that has repeated increasingly often in recent months: former research leaders from Google, OpenAI, or Anthropic start their own companies and close funding rounds worth tens of millions of dollars before they've shown any working technology.

For the AI startup market, it's a sign that capital is still flowing widely, despite growing questions about the profitability of many investments in the sector. Funds are betting that access to talent who helped build breakthrough models like Gemini is, on its own, sufficient protection against risk.

What it means for the industry

If Elorian's thesis holds up, closing the visual gap could matter well beyond chatbots and text assistants. Better visual reasoning is a prerequisite for robotics, autonomous vehicles, and systems that analyze medical images, where precisely recognizing shapes and spatial relationships matters more than generating fluent text.

For now, Elorian has not shown a working product publicly. The company's next test will be whether, in the coming months, it can turn its high-profile funding round and its founders' reputations into concrete research results that actually narrow the gap between how AI models handle language and how they handle images.