Out of the Lab, Onto the Line

Fudan Character

DING Wenchao has worn three hats in five years: Huawei "Genius Youth" building autonomous driving systems, tenure-track professor at Fudan probing the frontiers of embodied intelligence, and now co-founder and chief scientist of TARS, a startup that raised nearly 5 billion yuan in its first year.

This year, he added another line to his CV: a spot on MIT Technology Review's 2025 "35 Innovators Under 35" (TR35) China list.

From Huawei's "Genius Youth" to Researcher at Fudan University

DING Wenchao's research has always circled one question: how to make robots complete real physical tasks reliably through learning.

After earning his bachelor's degree from Huazhong University of Science and Technology in 2015, he began his doctoral studies at Hong Kong University of Science and Technology. But by his second year, he realized electronic information felt "invisible and intangible" to him. Robotics, by contrast, was something he could see, touch, and watch come alive. He decided to switch fields.

From drones to autonomous driving, and then to more general embodied intelligence, Ding explored different research directions. In particular, after tuning the algorithms for an autonomous vehicle, he would often be the first person to get in and test it. "It feels great when seeing things physically moving," he said.

In 2020, Ding joined Huawei's "Genius Youth" program as a technical expert in prediction for the automotive business unit. He contributed to Huawei ADS 1.0 and 2.0, proposing a prediction-decision coupling framework that made vehicles interact more actively with traffic participants. The technology eventually entered mass production and was applied to multiple models.

By 2023, autonomous driving had become a crowded, increasingly mature field. But to Ding, autonomous driving "had few secrets left." So he left Huawei and joined Fudan University. "At Fudan, the open atmosphere pushes me to think about more general and deeper questions," he said.

Co-founding TARS: Work-Ready, Not Just Demo-Ready

Ding is convinced that the value of embodied intelligence ultimately needs to be validated in real-world industrial settings. "Universities are good at exploring the unknown frontier, especially fundamental questions. When those questions are answered, or while much uncertainty remains, proven results often need to be scaled up to create real productive value for society," he said.

That conviction is what drove him to start TARS with former Huawei autonomous driving CTO CHEN Yilun and others, serving as co-founder and chief scientist. Rather than chasing flashy one-off robots, TARS takes a scalable approach to data collection for model training, and deploys these models across real production lines.

The core idea is straightforward: humans are already the world's most capable agents, so why not learn directly from them?

Ding proposed a human-centric data collection paradigm and an AI World Engine (AWE) embodied model, enabling robots to learn perception, decision-making, and manipulation by observing humans.

Under Ding's direction, TARS built SenseHub, a wearable device that captures human motion data in complex real-world scenarios cheaply and at scale. The work earned him a spot on MIT Technology Review's 2025 TR35 China list.

The market has taken notice. Within two months of founding, TARS closed a $120 million angel round, followed this April by a $455 million Pre-A-a record for China's embodied intelligence industry.

Choose Hard. Build for the Long Run.

Ding also serves as a PhD supervisor at Fudan. His approach to research was shaped early on by his own PhD advisor, who insisted he validate autonomous driving algorithms on real roads rather than simulations. That lesson stuck. Today, Ding holds his students to the same standard: if a result can't be demonstrated in the physical world, it doesn't count, no matter how elegant the theory.

His lab strikes a deliberate balance between rigor and freedom. Students are given wide latitude to explore, but they're expected to ground their ideas in reality. This semester, for instance, his machine learning course will give students access to simulators and real hardware to test classification and regression problems in physical settings, not just on a screen.

That same philosophy underpins TARS. Its name, a phonetic nod to the Chinese word 踏实 (tā shi), meaning "down-to-earth," reflects what Ding tells his students: in an age of endless options and quick wins, be the people who tackle the hard problems and do work that is valuable over the long term.

(END)

Writer: WEI Siqi

Proofreader: WU Zhengyang

Editor: LI Yijie