Selected work & papers
The question under all of it: how can embodied or resource-constrained agents build predictive internal models of the world, adapt continuously after deployment, and learn efficiently using local, predictive, or non-standard learning rules?
Predictive world models for robot manipulation
A compact, self-supervised predictive world model with continual learning across multiple time-scales, evaluated on the LIBERO manipulation benchmark. The question I'm chasing: can an agent keep adapting (at test time, across tasks, over its lifetime) instead of freezing the moment it ships?
CLP — Continually Learning Prototypes
A rehearsal-free continual learning algorithm for robots that meet new objects and environments after deployment. It cuts catastrophic forgetting by about 85% and adapts from a handful of labelled examples by reusing pretrained representations.
On-chip continual learning on Intel Loihi 2
A spiking version of CLP co-designed with neuromorphic silicon for sub-millisecond on-device learning. Against the strongest edge-GPU baseline it reaches 113x lower latency and 6,600x lower energy, while matching replay-based accuracy and staying rehearsal-free.
CLANE — action recognition on neuromorphic hardware
An end-to-end on-chip spiking pipeline that learns human actions from event-camera streams, class-incrementally, at about 5 ms per sample and millijoule-level energy: over 100x less energy and 16x lower latency than a Jetson Orin Nano baseline.
Cerebellum-inspired adaptive control on Loihi
A neural control model inspired by the cerebellum, deployed on Intel's Loihi chip for closed-loop control of a bio-mimetic robot arm: adaptation in the loop, running on neuromorphic hardware.
-
Online Continual Learning on Intel Loihi 2 via a Co-designed Spiking Neural Network
first author under review
-
CLANE: Continual Learning of Actions on Neuromorphic Hardware from Event Cameras
first author
-
Continual Learning for Autonomous Robots: A Prototype-based Approach
first author
-
Efficient Online Learning with Predictive Coding Networks
co-author
-
Interactive Continual Learning for Robots: a Neuromorphic Approach
first author Best Paper Award
Best Paper Award — ICONS 2022
Interactive continual learning for robots
Intel "Fearless Innovation" Award
2022
DAAD Scholarship
Master's studies, German Academic Exchange Service
Bernstein SMARTSTART Fellow
Computational Neuroscience