--- title: "Former Google DeepMind Researcher Launches AI Startup…" canonical: "https://www.metaintro.com/blog/former-deepmind-researcher-elorian-ai-startup-2026" language: "en" author: "drashtigarach" published: "2026-05-21T12:50:12.000Z" modified: "2026-10-02T19:29:36.584Z" --- [Back to Blog](/blog) [AI](/blog/tag/ai)[Hiring](/blog/tag/hiring)[Jobs](/blog/tag/jobs)[Technology](/blog/tag/technology)[News](/blog/tag/news)[Information](/blog/tag/information) # Former Google DeepMind Researcher Launches AI Startup in 2026 as Big Tech Brain Drain Grows A former Google DeepMind researcher just launched visual reasoning AI startup Elorian — the latest signal of Big Tech's AI talent brain drain in 2026 again. [![Drashti Garach](https://cdn.metaintro.com/rs:fill:40:40/q:72/plain/images/5719d740-e510-42bc-8017-e040d145f35f_1766029465094.png)Drashti Garach @DrashtiGarach](/blog/author/drashtigarach) [May 21, 2026](/blog/archive/2026/05)14 min read ![Former Google DeepMind Researcher Launches AI Startup in 2026 as Big Tech Brain Drain Grows](https://cdn.metaintro.com/rs:fill:1200:675/q:78/plain/images/kai.i7RxcJ1J.png) [https://x.com/intent/tweet?text=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026](https://x.com/intent/tweet?text=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026&title=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026&title=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows)[mailto:?subject=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026](mailto:?subject=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026) A new visual reasoning startup called Elorian is the latest chapter in a story that has quietly defined AI hiring for the past three years. [Bloomberg's interview with Andrew Dai](https://www.bloomberg.com/news/videos/2026-05-21/former-google-deepmind-researcher-on-new-ai-startup-video), co-founder and CEO of Elorian and a former Google DeepMind researcher, ran on the sidelines of the JPMorgan Global China Summit on May 21, 2026. The on-camera conversation covered the progress of the company, hiring intentions, and funding plans, the same trio of questions every senior researcher leaving a frontier lab is now asked the moment they hang out a shingle. What is striking is not that one more DeepMind alumnus is starting a company. It is that the cadence has not slowed. Through 2024, 2025, and into 2026, names that used to sit deep inside Google, OpenAI, Meta, and Anthropic keep surfacing on cap tables and pitch decks instead of paper bylines. The Bloomberg clip with Dai is a snapshot of a broader migration that engineers and operators should be reading carefully, because the talent flow is rewriting where the best AI jobs actually live. ## What we actually know about Andrew Dai and Elorian? The verified facts from the Bloomberg segment are thin, and that is on purpose. Dai is described as co-founder and CEO of Elorian, a visual reasoning AI firm. His background is in Google DeepMind. The interview was conducted by Bloomberg's Haslinda Amin from the sidelines of the JPMorgan Global China Summit, and the topics on the table were the progress of the new startup, hiring, and funding plans. That is the entire confirmed record at the time of publication. There are no public Elorian funding numbers tied to this appearance, no announced headcount, no confirmed lead investor, and no product launch date attached to the clip. Anything beyond those five facts will need to wait for Elorian's own disclosures or a follow-up story from Bloomberg, the Financial Times, or The Information. That thin record is itself part of the story. Founders who came out of DeepMind, OpenAI, and Anthropic tend to operate in stealth or near-stealth for a long stretch, then surface with a single high-leverage interview at a venue like JPMorgan's China Summit, Sequoia's AI Ascent, or the Cerebral Valley conference. By the time the press cycle starts, the team is largely built and the round is largely closed. If you are an engineer reading these stories hoping to apply, you are almost always reading them too late. ## The brain drain is not new, but it has changed shape? Senior researchers have left Google, Microsoft, and Meta for startups for as long as those companies have existed. What is different this cycle is the gravitational pull of the frontier labs themselves. For most of the 2010s, the prestige path ran the other direction. The best PhD students wanted offers from DeepMind, FAIR, Google Brain, and later OpenAI and Anthropic. Joining one of those labs was the destination. Starting a company was the consolation prize for people who could not get hired, or a quiet retirement project for senior staff who had already cashed out. That hierarchy inverted somewhere between the public launch of ChatGPT and the first wave of multibillion dollar AI infrastructure rounds. Once it became obvious that the underlying research was going to be commercialized in months rather than decades, the math changed. A senior research scientist with two years of frontier-lab experience could credibly raise a seed round at a valuation that would take a decade of Big Tech stock vesting to match. The opportunity cost of staying inside a lab became real, and visible, and quantifiable in a way it had never been before. The result is a steady drip of founder departures, of which Andrew Dai is the most recent public example. It is not a sudden exodus and it is not a collapse of the labs. It is a slow rebalancing of where the most ambitious AI talent thinks the next decade of value will be captured. ## Why visual reasoning is the subfield to watch? The phrase "visual reasoning" is doing a lot of work in the Elorian description, and it is worth slowing down on. Visual reasoning sits at the intersection of computer vision, multimodal large language models, and structured planning. It is the capability that lets a system look at a diagram, a screenshot, a chart, a UI, or a real-world scene and reason about it the way a human would, not just label what is in the frame. That capability has become commercially urgent for three reasons. First, enterprise customers want AI agents that can operate software the way employees do, which means reading interfaces visually rather than relying on brittle APIs. Second, robotics and autonomous systems need models that can plan over visual inputs in real time. Third, regulated industries like healthcare, insurance, and finance generate enormous volumes of visual documents that current LLMs handle poorly. Founders coming out of the frontier labs increasingly pick subfields with that profile, narrow enough to defend against the giants, deep enough to require frontier-lab pedigree, and adjacent to a large existing budget line. Visual reasoning fits cleanly. So do agentic coding, biology-specific foundation models, on-device inference, and AI for legal discovery. If you are an engineer trying to read where the next round of well-funded startups will hire, those are the categories to track. ## What is actually driving senior researchers out the door? The standard answer is equity, and equity is real. A founder or early employee at an AI startup that raises at a competitive valuation can build wealth on a timeline that no Big Tech compensation package matches, even after the recent rerating of large-cap tech RSUs. But equity alone does not explain the pattern, because senior researchers at Google, Meta, and Microsoft are already among the highest-paid technical employees in the world. Three other forces matter as much. The first is autonomy. Inside a frontier lab, even principal researchers have to negotiate compute allocation, publication review, productization timelines, and internal politics. At a fifteen-person startup, the founder decides what gets built on Monday and ships it on Friday. For someone who spent the last five years watching their best ideas get absorbed into a roadmap they did not control, that shift is intoxicating. The second is focus. Big labs work on everything, which means individual researchers often work on narrow slices of a sprawling agenda. A startup forces a single hard problem, which is what many senior researchers actually want after a decade of breadth. The third is timing. The window to build a defensible AI company on top of current foundation models is genuinely finite. Researchers who believe the window closes in two to three years are leaving now to make sure they are inside it before the dynamics shift again. That urgency is not marketing. It is showing up in every conversation with founders who left in the last twelve months. ## What this means if you are an engineer thinking about leaving Big Tech? Most engineers reading about a DeepMind founder departure are not principal researchers with a clean shot at a seed round. They are senior or staff engineers, product-side ML engineers, infrastructure people, and applied scientists who can see the trajectory but are not at the top of the pyramid. The right read for that audience is different from the founder-track read. If you are in that bucket, the brain drain is not a signal to quit and start something. It is a signal that the best non-founder AI jobs are increasingly at startups founded by people who left the labs in the last two to three years. Those companies are hiring at every level, they are paying competitively in cash, and they are giving out equity that has a credible path to value, not just a lottery ticket. The catch is that the bar is high and the screening is brutal. Founders who came out of DeepMind or OpenAI hire the way they themselves were hired, which means a lot of weight on first-principles problem solving, prior shipped work, and clear thinking under time pressure. Polishing your interview prep matters more than polishing your resume. It is also worth being honest about the failure rate. Most of these startups will not become the next OpenAI or Anthropic. Many will be acquired in two to four years for a price that makes the founders rich and rewards employees with a meaningful but not life-changing payout. A smaller number will fold quietly. Joining one is closer to a high-conviction bet than a safe job change, and it should be sized accordingly inside a broader career plan. ## How to negotiate compensation and equity at an AI startup? If you decide to make the jump, the negotiation looks different from a Big Tech offer. There are a few rules that experienced startup hires consistently use. Treat the equity grant as the headline number, not the salary. A standard early engineer grant at a Series A AI startup is meaningful in percent of the company, and the dollar value will depend almost entirely on the next two rounds. Ask for the current fully diluted share count, the most recent preferred price, and the strike price on your options. If any of those three numbers are withheld, that is information. Negotiate the vesting cliff and acceleration. Standard four-year vesting with a one-year cliff is normal. Single-trigger or double-trigger acceleration on a sale is not standard for early employees, but it is worth asking for, especially if the company is plausibly acquirable in the next twenty-four months. Cash compensation should still cover your life. Many AI startup offers come in below FAANG cash bands by design. That is acceptable, but only if your equity stake is large enough to compensate. If both numbers are below market, you are subsidizing the founders, and you should walk. Get the title clarified in writing. Founding engineer, early engineer, and member of technical staff all mean different things at different companies and carry different downstream signaling value. Two years from now, when you are interviewing somewhere else or starting your own thing, your title at a hot AI startup will matter more than you expect. ## What to actually watch next? The Elorian announcement will likely be followed by a more detailed funding story within the next three to six months. Watch for the lead investor, the round size, and the named hires. Those three data points will tell you more about the visual reasoning subfield than any product demo. Also watch the rate at which senior researchers continue to leave the major labs. If departures accelerate through the back half of 2026, the founder-startup ecosystem will get more competitive for talent and the bar for joining one will keep rising. If departures slow, it will signal that the labs have successfully repriced their internal compensation and equity packages to retain the people who matter most. Either outcome reshapes the job market for everyone downstream. For now, the Andrew Dai interview is one more data point on a line that has been pointing in the same direction for three years. The senior end of AI research keeps voting with its feet, and the companies they are starting are increasingly where the next wave of strong engineering jobs will live. ## How the brain drain reshapes Big Tech AI labs in 2026? The Big Tech AI brain drain has a second-order effect that is easy to miss from the founder side. Every senior researcher who leaves to start a company takes a piece of institutional knowledge with them, and Google, Meta, OpenAI, and Anthropic spend the next quarter or two trying to backfill not just the headcount but the judgment. The result inside the labs is more management overhead, more compensation packages tilted toward retention, and a slow shift in research direction toward whatever the remaining senior people care about. For engineers who stay, that can be a real opportunity. Promotion timelines compress when senior researchers exit on a regular cadence. Internal mobility opens up because empty seats need to be filled. The catch is that the work itself often shifts toward more applied, more product-focused problems and away from the open-ended research that originally attracted many of those engineers. Knowing your own appetite for that shift is one of the cleanest career decisions you can make right now, and it is the lens worth using whenever the next founder announcement hits the wire. ## How to read the next founder announcement before it hits the press? Founder departures from frontier AI labs do not arrive randomly. They tend to cluster around three windows. The first is the start of each calendar year, when annual stock vests, and senior researchers can leave without leaving large unvested equity on the table. The second is the end of major model release cycles, when the next research roadmap is becoming clear and people who disagree with the direction make their move. The third is the run-up to industry summits like JPMorgan Global China, Sequoia AI Ascent, or Cerebral Valley, where a short interview functions as a soft launch for the new company. If you are an engineer scanning the news for early signals, those are the three windows worth setting calendar reminders against. The other useful filter is the language a founder uses on first appearance. Founders who lead with the research problem they want to solve, like visual reasoning in Andrew Dai's case, tend to attract better engineering talent and more patient investors. Founders who lead with the size of the market or the speed of fundraising tend to attract a different mix that does not always survive the first real customer. Reading the framing of the first interview is one of the cleaner predictors of whether a new AI company is built to last five years or built to flip in eighteen months. For engineers thinking about which startup to join, that signal is worth more than any title or comp-band number. ## People Also Asked ### Q: Who is Andrew Dai and what is Elorian? A: Andrew Dai is the co-founder and CEO of Elorian, a visual reasoning AI startup. He is a former Google DeepMind researcher. He discussed the company's progress, hiring, and funding plans with Bloomberg from the sidelines of the JPMorgan Global China Summit in May 2026. ### Q: Why are so many AI researchers leaving Google DeepMind, OpenAI, Meta, and Anthropic? A: The most cited reasons are equity upside at startups, more autonomy over what gets built, focus on a single hard problem instead of a sprawling research agenda, and a belief that the window to build a defensible AI company on current foundation models is finite. Cash compensation at the labs is still strong, so the pull is about ownership and direction more than salary. ### Q: Is now a good time for engineers to join an AI startup founded by ex-Big-Tech researchers? A: For senior and staff engineers with relevant ML or infrastructure experience, the early-stage AI startup market is one of the most competitive hiring environments in years. The catch is a high interview bar, a high failure rate among individual companies, and equity packages that need careful evaluation. It is a high-conviction bet, not a safer alternative to Big Tech. --- ## Related Articles - [North Dakota Workers Face AI Job Displacement Risk](https://www.metaintro.com/blog/north-dakota-workers-ai-job-displacement-risk) - [General Motors Job Cuts Driven by AI](https://www.metaintro.com/blog/general-motors-job-cuts-ai) - [Tesla Robot Training Job Opens Door to AI Hardware Careers](https://www.metaintro.com/blog/tesla-robot-training-job) - [Software Engineer Machine Learning at Whatnot](https://www.metaintro.com/blog/software-engineer-machine-learning-whatnot) - [Senior Software Engineer Platform at Stronghold](https://www.metaintro.com/blog/senior-software-engineer-platform-stronghold) - [DeFi Innovation Engineer Internship at Keyrock](https://www.metaintro.com/blog/defi-innovation-engineer-internship-keyrock) - [Cisco Second Round of Layoffs 2024](https://www.metaintro.com/blog/cisco-second-round-layoffs-2024) - [Intel Layoffs 2024 and the Chip Talent Reshuffle](https://www.metaintro.com/blog/intel-layoffs-2024) --- Future-proof your career. Metaintro tracks AI hiring and founder trends, delivered to your inbox weekly. [Sign up free](https://www.metaintro.com). ### Share this article [https://x.com/intent/tweet?text=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026](https://x.com/intent/tweet?text=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026&title=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026&title=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows)[mailto:?subject=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026](mailto:?subject=Former%20Google%20DeepMind%20Researcher%20Launches%20AI%20Startup%20in%202026%20as%20Big%20Tech%20Brain%20Drain%20Grows&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fformer-deepmind-researcher-elorian-ai-startup-2026) ![](https://cdn.metaintro.com/rs:fill:1200:800/q:30/plain/images/bridges/bridge-expand.1df895c6bd76d96f.png) For job seekers ## Ready to find a role that actually fits? Upload your résumé, start a Job Search Thread, and let Metaintro rank real openings against your experience — then guide you from search to offer. [Get Started Free](/signup)[Search matching jobs](/jobs/search) Match Compare live roles against your current evidence. Position Turn proof projects into role-specific applications. Improve Use market feedback to keep the skill plan current. [Return to navigation](#main-navigation)