Teams at Modular

Design, implement, and optimize high-performance attention and foundational AI kernels for GPUs, NPUs, ASICs, and other accelerators. Build reusable attention-framework abstractions, profile workloads, develop benchmarks and correctness tests, analyze hardware and compiler behavior, and improve production inference performance. Collaborate with hardware and software teams, document designs, resolve ambiguous performance issues, and mentor engineers while supporting new accelerator architectures and AI workloads.