Maciej Besta

I lead research on sparse graph computations and large language models at the Scalable Parallel Computing Lab (SPCL) at ETH Zurich; I also work on network topologies and occasionally other aspects of the high-performance computing landscape. I also lead collaborations and project management at the SPCL's part of the ETH Future Computing Laboratory (EFCL). I am fortunate to work with and be mentored by Prof. Torsten Hoefler and to be a part of the rich and vibrant SPCL, EFCL, and ETH research environments.

I am particularly interested in irregular computations, such as graph neural networks, graph databases, graph streaming, sparse models, or classical graph algorithms, in any types of settings and workloads . I approach this problem holistically at different abstraction levels, including (1) hardware, (2) interconnects and their underlying network topologies, (3) middleware (runtimes, frameworks, OS), (4) algorithms, and (5) paradigms, abstractions, programming models.

The common effort in all these efforts is to understand better the underlying fundamental properties of a given problem, system, phenomenon, or solution. This is done with performance modeling or any other formal way of describing a given part of reality.