Who’s Leading XPeng Robotics: Four Key Figures Behind IRON

Behind XPeng's Sensational “IRON” Humanoid Robot

XPeng’s “IRON” robot stunned audiences at its launch event, walking with uncanny human-like steps — so realistic that some suspected a human was inside the suit. Here’s an inside look at the elite team driving this breakthrough.

---

Key Figures Leading XPeng’s Robotics Division

Among XPeng's robotics leadership, four public-facing names stand out, each with exceptional achievements.

Mi Liangchuan — VP of Robotics & AI Committee Chair

• Current Roles: VP of XPeng Motors Robotics, Chair of the AI Technology Committee, Project Lead for IRON.
• Responsibilities: Defines technology roadmaps, manages organizational structure, drives product implementation.

Innovation Highlight:

> “The new-generation IRON isn’t really building a ‘waist,’ it’s building a ‘spine.’”

The design mimics human anatomy — spine, muscle groups, abdominal cavity, and even skin — resulting in lifelike motion.

Turning Point:

In March 2023, after extensive training, IRON’s reverse walk suddenly became distinctly human-like, marking a breakthrough for XPeng’s generative controller.

Technical Enhancements:

• Innovative spine-inspired waist design
• Additional forefoot freedom degrees
• Proprietary generative controller

---

Career Path

• 1999: Admitted to USTC, Electrical Engineering (high-speed data acquisition).
• Graduate: Master’s in Robotics, Iowa State University (specializing in geometric fitting & dynamic control).
• Carnegie Mellon Robotics Institute: Electrical Engineer — obstacle avoidance for medical robots, laser drone navigation.
• 2005–2015 (NVIDIA):
• Software Engineer (OpenGL on Mac, GPU integration)
• Senior Software Engineer (GPU/SoC graphics, Tegra OTA cloud architecture — core to NVIDIA DRIVE)
• Senior Software Engineering Manager (managed Android software teams across US/China).
• 2015–2016: Co-founded Miwen Power (CTO).
• 2016–2021 (NVIDIA Return): Autonomous driving division — deep learning optimisation, performance tuning.
• 2021: Joined XPeng — Senior Director of Autonomous Driving.
• 2023: Head of Robotics; later VP & AI Committee Chair.

---

Chen Jie — Reinforcement Learning Lead

• Joined XPeng Robotics July 2025 to lead reinforcement learning efforts.
• Background:
• Bachelor’s — University of Hong Kong
• PhD — Harvard University (AI, reinforcement learning)
• Previous roles: ByteDance Seed team (Reinforcement Learning Head), Momenta (autonomous driving unicorn).
• Reasons for joining:
• XPeng’s early humanoid robot R&D foundation.
• Strong funding (~50 billion RMB reserves due to EV sales).

---

Ge Yixiao — Intelligent Mimicry Department Director

• Founder & Principal Researcher of XPeng’s Intelligent Mimicry Department.
• Bachelor’s — Huazhong University of Science and Technology
• PhD — MMLab, Chinese University of Hong Kong (multimodality research).
• Achievements: 10,000+ Google Scholar citations, technical leadership at Tencent ARC Lab (T12 Technical Expert by age 28), double Tencent Technical Breakthrough Award.

---

Liu Xianming — Head of XPeng Autonomous Driving Division

• Strong collaboration between autonomous driving and robotics teams.
• Focus: Eliminating language as an intermediate layer in Vision-Language Alignment (VLA) systems to reduce data loss and accelerate model inference.
• Academic background: PhDs from Harbin Institute of Technology and University of Illinois Urbana-Champaign; visiting scholar at Tsinghua and University of Texas at San Antonio.
• Industry experience: Microsoft Research Asia, Baidu IDL, Adobe, Facebook (global population density from satellite imagery), Horizon Robotics, Cruise (embedded AI, perception systems).

---

Early Talent Moves — The Doggo Robotics Acquisition

In 2020, XPeng acquired Doggo Robotics, one of China’s earliest quadruped robot developers.

• Founder Zhao Tongyang created a first-gen robot dog in under two months.
• He Xiaopeng envisioned quadruped robots as one of XPeng’s next growth pillars (alongside flying cars).
• Talent war: Salaries up to RMB 4 million/year attracted over 300 experts.

Challenges:

• Slower research cycles within corporate structures
• Internal debate on humanoid robot viability

April 2023: Zhao formed a small agile team (~20 ppl), producing XPeng's first humanoid — PX5, precursor to IRON.

Eventually Zhao left to found ZQQ Robotics, shrinking XPeng’s robotics team to ~70 members before Mi Liangchuan rebuilt the division past 200 members with new hires like Chen Jie & Ge Yixiao.

---

XPeng’s Strategic Shift

The company has now officially designated humanoid robots as its third growth curve after smart EVs and flying cars — paralleling Tesla’s approach.

Tesla Parallels

• XPeng founder He Xiaopeng openly inspired by Tesla’s philosophy.
• Design similarities noted in XPeng’s P7 model, mirroring Tesla’s Model 3 features.
• Shared vision strategies in autonomous driving.

---

Flying Cars & Market Dynamics

• XPeng’s flying car entered trial production (Nov 2024), with ~5,000 pre-orders; delivery scheduled for 2026.
• IRON’s debut brought XPeng deeper into “China’s Tesla” comparisons.
• Elon Musk himself acknowledged Chinese firms as formidable humanoid robot competitors.

---

Market Capitalization Comparison (Nov 2024)

• XPeng: ~$21.3 billion
• Tesla: ~$1.43 trillion (67x XPeng’s market cap)

---

References

• https://mp.weixin.qq.com/s/-v0A7Q1f8EHEp_WA5sb6jA
• https://eu.36kr.com/zh/p/3433980784938631
• https://www.sohu.com/a/952339997_121954550
• https://geyixiao.com/
• http://gairdao.com/app/articles/sxxgrbae
• https://www.thepaper.cn/newsDetail_forward_11631630
• https://www.linkedin.com/in/lc-liangchuan-mi-8058995/

---

AI Tools for Creators Tracking Robotics & AI Trends

For researchers, innovators, and creators exploring AI-powered robotics or autonomous systems, AiToEarn offers:

• Open-source ecosystem for AI content creation, publishing, monetization.
• Multi-platform distribution (Douyin, Bilibili, YouTube, X).
• Analytics and AI model rankings: AI模型排名.

AiToEarn’s integrated workflow mirrors XPeng’s own cross-domain collaboration — helping creators efficiently produce, publish, and monetize AI-driven insights at global scale.