Axel Feldmann
Axel Feldmann is a PhD candidate in electrical engineering and computer science whose research aims to address hardware-level performance bottlenecks in many commonly used numeric algorithms. As a MathWorks Fellow, Axel will expand his successful work creating hardware accelerators specifically designed for sparse linear algebra. Axel has already designed two accelerators—Spatula and Azul—that are over 200× faster than CPUs and GPUs on linear solvers. The efficiency and scalability of these systems enable transformative speedups; a single computer system using Axel’s accelerators and consuming under 2 kilowatts achieves comparable performance to supercomputers with over 100,000 cores consuming megawatts of power. Axel is now generalizing these designs to produce a more flexible and programmable architecture capable of speeding up more than just linear solvers. By deploying this hardware in the MathWorks cloud, millions of MATLAB and Simulink users could have access to the same compute power that today is only available to institutions with supercomputers. Thus, Axel’s work has the potential to democratize high-performance scientific computing and enable revolutionary gains in performance through hardware acceleration.