Less than two months after walking away from his high-profile gig at Huawei, former Noah’s Ark Lab director and Pangu LLM lead Wang Yunhe is already making waves. His freshly minted AI startup, Jiyuan Lüdong, just wrapped up its seed funding round, with a staggering $100 million post-money valuation backing it up. Word on the street? The clock was ticking fast.
According to official business filings, Shanghai Jiyuan Lüdong Technology Co., Ltd. officially registered on April 8, 2026, with Wang Yunhe taking the helm as both CEO and legal rep. He’s not going solo, though. Joining him as co-founder and CTO is Han Kai, a former lead researcher at Huawei’s Noah’s Ark Lab. Han comes stacked with credentials—from Zhejiang University to Peking University and the Chinese Academy of Sciences—and has spent years knee-deep in foundation models and deep learning. Fun fact: he and Wang were actually teammates back in the Huawei trenches. Now, they’re channeling that combined brainpower into Jiyuan Lüdong, laser-focused on the absolute firehose of talent and investment right now: AI Agents.
Born in 1991, Wang Yunhe quickly earned himself the nickname “the young commander of Pangu” within the Chinese tech scene. His academic roadmap reads like a dream for any AI researcher: a math undergrad at Xidian University, followed by a PhD in Intelligent Science from Peking University in 2018. There, he cut his teeth under some heavy hitters like Prof. Xu Chao and Australian National Academy Academician Prof. Tao Dacheng, zeroing in on everything from deep learning and computer vision to the critical art of model compression. But how did this academic powerhouse end up at Huawei? It started way back in 2017. While still grinding through his doctorate, Wang landed an internship at the Noah’s Ark Lab—actually making him the very first intern hired for Huawei’s Beijing branch. Fast forward to 2018, graduation day rolled around, and he signed his full-time contract. That moment kicked off what would become a solid eight-year chapter with the company.
While embedded in Huawei’s ecosystem, Wang didn’t just sit around. He helped pioneer GhostNet—a lightweight neural network architecture designed specifically for edge devices that seriously shook up the developer community. His chops were further validated when he was named a lead inventor on a breakthrough paper about highly efficient neural architectures for multiplying and adding operations, which ultimately landed on Huawei’s prestigious Fourth Annual “Top Ten Inventions” list. If you dig into his publication record, it’s pretty staggering: over 70 peer-reviewed papers across top-tier global conferences and journals, plus a Google Scholar citation count blowing past 33,000. Yeah, he’s got the receipts.
By March 2025, the pressure was on, and Wang stepped up to take the reins as director of the Noah’s Ark Lab, succeeding Yao Jun. Suddenly, he was steering the ship for the entire Pangu Large Language Model pipeline, cementing his spot as one of the absolute pillars of Huawei’s AI engineering team. He wasted no time getting his hands dirty. By June 2025, he’d shepherded out Pangu 5.5, a major milestone featuring a cutting-edge Mixture-of-Experts (MoE) architecture. Then, just recently on March 20, he stood front and center at Huawei’s massive press conference, officially announcing that the entire Pangu ecosystem was being thrown open to the public. Open-source move? Huge deal.
But every good chapter has to close eventually. Right when Pangu was hitting its stride, Wang took to WeChat Moments on March 28 to drop the bombshell: he was moving on. His sign-off message was surprisingly heartfelt—“Eight years down. Bittersweet to say goodbye to the place where it all happened.” Those words quietly wrapped up a nearly nine-year saga with Huawei. Naturally, the tech world went into guessing mode. Most insiders think it boils down to two things: a growing clash over the company’s long-term technical roadmap, and Wang simply feeling it was time to chase his own next big leap.
Here’s the thing nobody really talks about: Wang’s exit didn’t happen in a vacuum. Back in August 2025, just weeks before he packed his desk, Huawei Cloud decided to shake things up big time. They issued an internal directive that basically gutted and merged dozens of units under the Cloud Business Unit. What came out the other side? A leaner, meaner operation hyper-focused on a specific “3+2+1” framework covering general compute, AI compute, storage, AI PaaS, databases, and security. Naturally, chunks of the Pangu model division got folded into other teams. The industry buzz was loud: “Huawei is pulling back from non-core AI projects.” Huawei Cloud fired back pretty quickly, calling those claims total misreadings of the situation and doubling down that they’re actually pouring more resources into Pangu and smart compute infrastructure. So, why leave then? Maybe the timing was just right for him to strike out on his own.
To be fair, Wang isn’t jumping off the Huawei train because he’s crazy. He’s riding a massive wave. Over the last couple of years, we’ve watched at least a dozen of Huawei’s brightest minds in AI, self-driving tech, and custom silicon pack up and launch their own ventures. And where are they all landing? Exactly where the market’s screaming for talent: AI Agents, embodied intelligence, and next-gen autonomous systems. Look at the track record—it works. Former self-driving lead Su Qing built out a fully funded AV startup, and the guy who ran Huawei’s Ascend chip division dropped his own AI silicon company that blew past typical seed checks. When you’ve been building at this scale, striking out feels less like a risk and more like the obvious next step.
So, what’s the actual play here? Jiyuan Lüdong is putting all its chips on the table for AI Agents, a sector widely acknowledged as the holy grail sitting right behind foundational LLMs. Think of it this way: while traditional models just spit out text, an AI Agent actually does stuff. It figures out what you need, maps out the steps, grabs the right APIs or tools, runs the workflow, and learns from the results—all completely hands-free. In messy, real-world environments, agents handle end-to-end workflows that would normally tie up a whole team. It’s the difference between having a brilliant encyclopedia and hiring a proactive junior associate who actually gets things done.
The numbers back up the hype. Back in 2025 alone, China’s AI Agent market exploded past the 18.2 billion yuan mark—a jaw-dropping 78% jump year-over-year. We’re firmly in the breakout phase. You’ve got the usual suspects lining up: Microsoft, Google, Baidu, Alibaba… everyone wants a slice of this pie. But don’t let the giants scare you; it’s also an absolute gold rush for scrappy startups. The room is getting crowded fast, which means speed, sharp execution, and a genuinely useful product are going to separate the winners from the noise.
So how do they plan to win in such a packed arena? Jiyuan Lüdong is targeting the enterprise market head-on with a universal AI Agent platform. And here’s where Wang’s background really shines: they’re leaning hard into his deep expertise in model compression. Instead of forcing companies to burn through millions on raw GPU clusters, they’re engineering sleek, lightweight agent frameworks that run efficiently without breaking the bank or requiring a dedicated DevOps army to manage the deployment. Basically, they’re tackling the two biggest headaches in enterprise AI today: sky-high inference costs and ridiculously complicated setup processes. If they pull it off, they’re solving a problem every CTO is currently stressing over.
