Jan Robert Finkbeiner

I hold a Bachelor's and Master's degree in physics, with a focus on computational physics and machine-learned potentials. At PGI-15, I investigate efficient computing approaches at the intersection of algorithms and hardware. I'm particularly interested in enabling intelligence at the edge, aiming to build adaptive and autonomous systems that can operate independently and efficiently. My research spans both algorithm design and hardware-aware implementation. In the area of brain-inspired architectures, I’ve worked with Spiking Neural Networks to explore their potential for energy-efficient inference and training. I’ve used platforms like Intel’s Loihi 2 and Graphcore’s IPU to take advantage of their inherent activation sparsity. More recently, I’ve been working on adapting well-established models such as transformers to in-memory computing platforms. This effort highlights how near- and in-memory computing can help meet the growing need for low-latency, low-power deployment of large language models.