Few things excite hardware investors more than founders with ties to DJI. And in artificial intelligence, a comparable signal has emerged through affiliation with SenseTime.
Drive along Shanghai’s Middle Ring Road and SenseTime’s sprawling headquarters rises above the overpass, its bulk resembling an aircraft carrier. A short walk down the tree-lined street behind it brings you to another AI bellwether, MiniMax.
The proximity is more than geographic. The two companies are closely linked by people and history. MiniMax founder Yan Junjie previously served as vice president, deputy head of research, and CTO of SenseTime’s smart city business unit.
MiniMax, along with fellow AI startup Zhipu AI, known internationally as Z.ai, has recently cleared its Hong Kong Stock Exchange hearings and released its prospectus. Both are expected to go public as early as January 2026.
With a business model that spans foundational models and consumer-facing applications, MiniMax is among a small group of Chinese AI companies moving steadily toward meaningful revenue. Its prospectus shows that in the first nine months of 2025, the company generated USD 53.4 million in revenue, surpassing its full-year 2024 total of USD 30.5 million. The filing also notes that its consumer products are nearing breakeven.
MiniMax is not the only notable company to emerge from SenseTime’s orbit.
One of the fastest-growing AI companies globally is Vivix AI, founded overseas by Liu Yu, SenseTime’s former executive research director and head of its AI-generated content (AIGC) division. Within ten months, while still in early-stage model development, Vivix’s valuation climbed to USD 1.32 billion.
Across nearly every major AI subfield, SenseTime alumni are increasingly in the headlines for launching startups of their own.
For industry insiders, the appeal is straightforward. It reflects both the visible success of earlier spinouts such as Momenta, founded by former SenseTime R&D executive director Cao Xudong, and MiniMax itself, as well as a shared background among founders that combines deep technical training with experience deploying AI at scale.
Scarcity as an edge
“Scarcity.” That is the word investors who have met Liu Yu most often use to describe him.
In AI entrepreneurship, scarcity refers to founders who combine a strong technical record with hands-on experience building real products. When Liu launched Vivix overseas in early 2025, investors including IDG Capital and HongShan moved quickly. According to a person familiar with the matter, one firm’s internal assessment concluded that fewer than five people globally in computer vision possessed both advanced algorithmic expertise and large-scale deployment experience. Liu, who led a 100-person foundational model team at SenseTime and managed more than 4,000 GPUs, was identified as one of them.
Many SenseTime alumni share similar profiles. Yan holds a PhD from the Chinese Academy of Sciences and previously led a 700-person team at SenseTime, helping advance its facial recognition algorithms. Luo Yuchen, founder of Yantu Technology, worked on China’s first dynamic facial recognition system for police use and delivered multiple large-scale algorithm deployments.
This talent pipeline draws directly from SenseTime’s longstanding strengths in natural language processing and computer vision. Several industry observers said that the company’s mature engineering pipelines and deployment systems, built during the previous AI cycle, could be adapted quickly for large model development. That advantage was especially valuable in 2023, when model architectures were still poorly understood.
An early MiniMax employee echoed that view, saying the company borrowed extensively from both SenseTime and Baidu, where Yan also worked, when building its technical teams and evaluation frameworks. “Even down to how we structured data labeling systems,” the person said. “Big companies have proven templates that can be deployed immediately.”
Technical credentials alone, however, are no longer enough to secure investor confidence. What distinguishes the latest wave of SenseTime alumni is their experience with products and commercialization.
At SenseTime, Liu led development of the AIGC product SenseMirage, which launched in September 2023. Within nine days, it surpassed three million users and reached 530,000 daily active users. With Midjourney unavailable in China at the time, SenseMirage filled a clear market gap.
That experience gave Liu not only a command of multimodal model development but also a clear understanding of distribution and user adoption. According to one Vivix investor, Liu’s edge lies in his focus on real-time interactive video, with rapid iteration across models and infrastructure contributing to the company’s valuation growth.
MiniMax’s head start
The capital market’s enthusiasm for SenseTime alumni is closely tied to MiniMax’s early momentum.
From its earliest days, MiniMax made several bets ahead of the curve. In 2023, while much of the industry focused on model size and benchmark performance, the company already operated a globally popular AI app, Talkie. In January 2024, months before mixture-of-experts (MoE) architectures gained broader attention, MiniMax released abab 6, which is often described as China’s first large model built on an MoE architecture.
By the second quarter of 2024, MiniMax shifted its focus to multimodal AI. Subsequent releases, including its speech model series and the AI video generator Hailuo AI, have consistently ranked among leading products in both domestic and overseas markets.
With multiple products and diversified revenue streams, MiniMax weathered the DeepSeek-driven industry shakeout with limited disruption.
Reapplying proven patterns is a familiar way for investors to manage risk, and Yan’s track record at SenseTime made the “SenseTime alumni” label particularly compelling. Liu, now viewed by many investors as a breakout founder of 2025, had long been considered a likely startup leader. A person familiar with the matter said that when IDG Capital sought to back an overseas-focused multimodal AI company, with MiniMax already covering the domestic market, Liu’s background made him the natural choice.
Another example is Vast, a 3D generation startup founded by Song Yachen. Song previously worked in SenseTime’s CEO office and AI gaming division before joining MiniMax as a co-founder. He left in late 2022 to launch Vast.
At the time, 3D generation attracted limited attention. Backed by his SenseTime and MiniMax credentials, Vast quickly reached a USD 200 million valuation. Today, it is widely regarded as one of the leading startups applying AI to 3D modeling.
Even so, some investors are beginning to pull back. “Some projects are being valued on par with major model developers, without real products or data,” one investor told 36Kr, adding that this “runs counter to basic market logic.” He compared the current moment to 2024, when investors rushed to back former ByteDance employees, only for many projects to underperform a year later.
Building a company, after all, rarely hinges on a single factor. In a recent podcast hosted by Luo Yonghao, Yan reflected on MiniMax’s trajectory, pointing to early tradeoffs, a clear view of scaling laws, and an organizational model that rejected traditional performance evaluation systems.
As one AI practitioner who listened to the podcast put it:
“You can copy someone’s resume, but you can’t copy Yan Junjie.”
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Zhou Xinyu for 36Kr.