Matei Coldea

Matei Coldea is a Yale junior studying Computer Science and Electrical Engineering. Over the past four years he has worked on robotics and embedded systems: at Accenture as an intern, then at Lenovo, where he built CSI-based wireless sensing systems combined with AI. He founded Verlet Robotics, where he built an advanced robotics lab from the ground up in East Africa, a region with no existing infrastructure for robotics or AI research, and where the team now collects teleoperation, UMI, and egocentric data at scale. On top of that he built the full data processing pipeline: motion retargeting, annotation tooling with reusable primitives like MANO hand-pose correction, and the operational software that keeps the whole thing running. With the Apollo Lab at Yale he has co-authored several papers spanning robot sensing, batch sampling methods, and most recently a CoRL 2026 paper on segmenting what a robot attends to using its attention matrices and task objective, aimed at improving VLAs and world models. He is also one of the main contributors to the MathWorks-sponsored Neuron project, where he built advanced visualization and approximation methods for neuron analysis.