OpenAI offers 3.85 million annual salary to expand its robotics team! Recruiting basic model talents, Peking University alumni are founding members

Written by
Caleb Hayes
Updated on:July-14th-2025
Recommendation

OpenAI is recruiting robotics talents with high salaries, and Peking University alumni are joining the team to explore cutting-edge technologies.

Core content:
1. OpenAI's robotics team is restarting, offering up to $530,000 in annual salary to recruit talents
2. Team goals: Develop the underlying technology model of robots and expand functions
3. Peking University alumnus Lin Xingyu, a founding member of the team, has a rich academic and research background

Yang Fangxian
Founder of 53AI/Most Valuable Expert of Tencent Cloud (TVP)

With an annual salary of up to $530,000, the OpenAI robotics team is hiring again!

The latest news is that OpenAI is recruiting robot underlying technology model developers with an annual salary of US$295,000 to US$ 530,000 (approximately RMB 2.14 million to RMB 3.85 million) .

The person who made this news known was Lin Xingyu , a founding member of the restarted OpenAI robotics team and a Chinese PhD from Carnegie Mellon University .

But a month ago, OpenAI’s recruitment notice was posted on LinkedIn, which happened to be very close to the time when Figure announced its “breakup” with OpenAI .

After the breakup, Figure also moved quickly. In just half a month, it released its own robot control model Helix, and then released a new robot Figure 02.

So some people speculated in Lin Xingyu's comment section that this time OpenAI is likely to create a model like Helix.

01

OpenAI wants to build a basic model for robots

OpenAI's JD said the team's work is to "extend the capabilities of the base model to support general-purpose robots in dynamic real-world environments, ensuring that they operate reliably and safely."

Specific tasks involved include but are not limited to action generation, motion planning, world modeling, and real-time voice conversations with emotions.

The main job of the recruits is to develop basic models for robots for OpenAI and help determine future robot development plans.

At the end of May last year, Forbes and The Information both broke the news that OpenAI's disbanded robotics team would be restarted, but in the news, OpenAI's goal was still to provide support to its partners.

It was not until August that the OpenAI robotics team was truly restarted, and Lin Xingyu, mentioned at the beginning, was a founding member of the team.

Lin Xingyu graduated from Peking University's School of Computer Science , and then studied for a doctorate under Associate Professor David Held at Carnegie Mellon University on the topic of "Learning Structured World Models for Manipulation of Deformable Objects". He will graduate in 2022.

During his doctoral studies, Lin Xingyu also interned at NVIDIA's robotics laboratory and the MIT-IBM joint laboratory.

After graduating with a Ph.D. and before joining OpenAI, Lin Xingyu conducted postdoctoral research under Pieter Abbeel , a machine learning expert at UC Berkeley and Andrew Ng’s first Ph.D. student .

His research interests are still the structured world models he worked on during his doctoral studies, but they are no longer limited to deformable objects. Instead, they have the ability to abstract time and space. His ultimate goal is to build a general robot that can assist humans in unstructured environments.

Back to OpenAI, there was another wave of recruitment at the beginning of this year that was also related to robots, but the main work content was more hardware-oriented.

The person who announced this at X at the time was Caitlin Kalinowski (former head of Meta's robotics and consumer hardware team) , a technician poached by OpenAI from Meta , and President Brockman also forwarded it.

The recruitment at the beginning of the year was seen as a landmark move by OpenAI to enter the field of robotics hardware. Now the company is also recruiting people in the basic model field, indicating that OpenAI's robots will begin to be fully self-developed.

02

Is comprehensive self-development the only way for OpenAI robots?

The breakup between OpenAI and Figure happened suddenly, and many people were surprised. At that time, the cooperation between OpenAI and Figure had just started for less than a year.

At that time, Figure founder Brett Adcock said that there had been a major breakthrough in internal end-to-end robotic AI and that he would demonstrate "something no one has ever seen on a humanoid robot" in the next 30 days .

At that time, Figure also planned to raise funds at a valuation of US$39.5 billion, which was a 15-fold increase compared to its previous valuation.

However, just over half a month after the breakup, Figure announced the launch of the end-to-end universal control model Helix, and then the Helix-driven robot Figure 02 also came into the view of netizens.

Helix was “something that no one had ever seen in a humanoid robot” and was the reason Figure dared to break up with OpenAI.

The CEO also revealed that this work took them more than a year and was aimed at solving general robotics problems, which means that Figure was already working on its own models before working with OpenAI.

Ignacio de Gregorio, a columnist for Medium, analyzed that 2025 will undoubtedly be a year in which robots will develop at a crazy pace, but at the same time, as the way of building AI skeletons becomes more popular, robot manufacturers' dependence on basic model suppliers such as OpenAI will become lower and lower .

And given the high price of the OpenAI model, it is not easy to make money on the robot by exporting the model.

Therefore, although the "broken up" OAI still has two robotics companies, 1X and Physical Intelligence, if it wants to steadily advance its own robotics plan, comprehensive self-research may be the safest way for OpenAI.

After all, Physical Intelligence also has its own model π0.

Judging from a series of actions since the restart of the robot plan was exposed, OpenAI seems to have realized the importance of comprehensive self-research.