At Config, we are building the data and foundation models for general-purpose robots that can perform useful two-handed tasks in real-world settings.

A central challenge in this effort is not just learning from data, but how that data is generated — specifically, how interactions are defined, captured, and executed in the physical world.

This is where the end-effector — the part of the system that directly interacts with objects and the environment — becomes critical. It determines how interactions with the world are executed and captured as data, serving as a common interface across different embodiments, including both robotic and human-operated interfaces.

Designing this interface is fundamentally a dual problem: the end-effector must be effective for real-world manipulation, while also producing high-quality, learnable data for training robot foundation models. These requirements can be in tension, and resolving that tension is a core part of the problem we are solving.

As a Mechanical Engineer, you will design and build robot end-effectors and/or human-operated grippers used for large-scale data generation. Your work will directly shape how robots learn, by defining how interactions are represented, captured, and executed in the physical world.

This is a unique opportunity to work at the interface between mechanical systems and AI — where hardware design directly determines the quality of data and, ultimately, the capability of robot foundation models.

Most robot learning systems are bottlenecked not by models, but by the quality of interaction data — and that quality is defined at the interface between humans, robots, and the physical world.

In this role, you will define that interface.

If you’re interested in building the physical foundation for robot intelligence — and seeing your work directly translate into real-world capability at scale — we’d love to hear from you.