Robot foundation models
How can broad pretrained models become useful, grounded components of robotic systems rather than isolated predictors?
Research
Research in progress.
My doctoral research is still taking shape. I’m exploring how robots can combine broad learned representations with adaptive decision-making in the physical world.
Current direction
Areas I’m currently exploring.
My focus is on learned systems that connect perception, uncertainty, and action in the physical world.
Active inference
How can perception, uncertainty, and action be treated as a connected process for adaptive decision-making?
Robotics in changing environments
How can learning systems remain reliable when they move from curated datasets into changing physical environments?
Current stage
I’m currently developing the research proposal, narrowing the central problem, and defining what a rigorous evaluation should look like.
This page is intentionally concise while that work is underway. As the research becomes public, it will grow to include clear problem statements, methods, experiments, and supporting artifacts.
What informs it
Autonomous inspection
Experience building and evaluating computer-vision systems for industrial inspection shaped my interest in reliability outside the laboratory.
Robotic sensing and edge AI
Working across sensors, embedded computing, and learning systems made the connection between perception and action concrete.
Reliability under change
Evaluating models across changing equipment, signals, and physical conditions made robustness a central part of the research question.