The whole picture
I'm an AI researcher based in Munich, finishing a PhD at LMU and TUM on continual learning for edge and embodied systems. For six years at Intel Labs' Neuromorphic Computing group I built real-time learning systems for autonomous agents, co-designing the learning algorithm with the hardware it runs on.
The through-line of my work is simple to state: post-deployment adaptation is the missing capability of edge AI. Most deployed models only ever run inference. I build agents that keep learning locally, on-device, under strict latency, energy, and memory budgets — measurable wins like sub-millisecond on-chip learning at thousands of times lower energy than an edge GPU.
Where I want to take this is the bridge between embodied AI and adaptive AI, with neuromorphic and edge computing as the moat behind it rather than the headline. My sharpest specialization is drone-class onboard autonomy — civil applications first: infrastructure inspection, agriculture, mapping, search and rescue, environmental monitoring.
Neuroscience
How brains learn online, consolidate, and forget — the original adaptive system, and still the best one.
Psychology
Perception, memory, attention, motivation: the behavioural layer that any theory of intelligence has to explain.
NeuroAI
Where neuroscience and AI inform each other — local learning rules, predictive coding, metaplasticity.
Philosophy of intelligence
What learning, agency, and understanding actually are, beneath the benchmarks and the hype.
Photography
Mostly travel and street; an excuse to look more carefully at light and structure.
Travel
New cities, long walks, and the small recalibration that comes from being somewhere unfamiliar.
Coffee
Treating extraction as a controlled experiment: variables, measurements, the occasional good cup.