Palona AI is building the operational intelligence layer for Physical AI.

A core part of our research is interaction understanding: treating interactions, rather than objects alone, as a first-class perception target. Our goal is to enable agentic systems that are not merely instruction-reactive, but event-proactive—systems that understand what is happening in the physical world, plan the right response, and deploy timely action.

We are now extending this capability into a closed-loop robotic system:

Perception → Interaction Understanding → Planning → Deployment → Execution Monitoring → Replanning

We are looking for a PhD Research Intern with a strong robotics background to help prototype and evaluate this next stage of our Physical AI stack.

The internship will focus on research and pre-production experiments that connect our interaction understanding models with real robots in dynamic environments. Initial use cases will span restaurants and healthcare, where timely, context-aware deployment decisions are critical.

This is a hands-on research role for someone who wants to work on problems that sit between frontier AI research and real-world robotics deployment.

You will have the opportunity to work on systems that are not limited to benchmark evaluation or simulation, but are designed to operate in real physical environments where perception, timing, reliability, and execution all matter.

What You’ll Work On

• Extend our interaction understanding models to support real-time robotic deployment decisions.
• Build closed-loop systems that connect perception, planning, robot execution, monitoring, and replanning.
• Develop methods for deciding when, where, and how a robot should act based on physical-world events and operational context.
• Integrate and program real robotic platforms for research and pre-production experiments.
• Monitor robot execution, detect failures or deviations, and trigger rapid replanning and redeployment.
• Design experiments for constrained but economically useful robotics tasks in restaurant and healthcare environments.
• Evaluate system robustness across changes in actors, objects, locations, timing, and environmental conditions.
• Explore how interaction-centric representations can support more reliable and generalizable robot behavior.
• Contribute to research publications, prototypes, and potentially patentable system innovations.

Requirements

What We’re Looking For

• Currently pursuing a PhD in Robotics, Computer Science, AI, Computer Vision, Machine Learning, or a related field.
• Strong hands-on robotics experience, ideally through work in an academic or industrial robotics lab.
• Experience programming, training, integrating, and debugging real robots.
• Familiarity with robotic perception, planning, control, or embodied AI systems.
• Strong software engineering skills, preferably in Python and C++.
• Experience working with modern deep learning frameworks such as PyTorch.
• Ability to move fluidly between research ideas and working physical systems.

Preferred Qualifications

• Publications at top AI, computer vision, or robotics venues such as NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, RSS, CoRL, ICRA, or IROS.
• Experience with vision-language models, video understanding, interaction recognition, embodied AI, or robot learning.
• Experience with real-time perception and decision-making systems.
• Experience with ROS / ROS2, simulation environments, motion planning, or robotic middleware.
• Experience deploying models or robotics systems outside of purely simulated settings.
• Background or domain understanding in healthcare, assistive robotics, hospital operations, elder care, or clinical environments.
• Experience working with limited, noisy, or weakly supervised data.