RoboForce AI just opened applications for their AI Residency program focused on embodied intelligence and real-world robotics.

Program specs:

• 3-6 month full-time commitment

• $10k/month compensation

• Access to large-scale GPU clusters and production infrastructure

Technical focus areas:

• Vision-Language-Action (VLA) models - multimodal architectures that map visual and language inputs directly to robotic control actions

• World models - learning predictive representations of environment dynamics for planning

• RL in physical systems - dealing with partial observability, sample efficiency, and sim-to-real transfer

• Real-world robot learning - handling distribution shift, safety constraints, and continuous adaptation

This is aimed at early-career researchers who want to work on the full stack from perception to control in physical environments, not just simulation. The interesting part here is they're explicitly calling out production-grade infrastructure, which suggests they're past the pure research phase and working on deployable systems.

For anyone working on embodied AI or wanting to transition from pure ML research to robotics applications, this could be a solid opportunity to see how VLA architectures and world models perform when they actually have to interface with messy physical reality.