Design and extend actor-like programming notations and compiler toolchains for an event-driven graph accelerator. Implement code generation, register and locality management, data-movement optimizations, and performance profiling. Maintain and extend compiler frameworks (LLVM/GCC/MLIR) and optimize for parallelism, memory hierarchy, and specialized hardware.

Lead design and implementation of a petabyte-scale unified virtual memory and system software stack for accelerator-based large-scale clusters. Build custom allocators, runtimes, schedulers, and Linux kernel/driver components. Integrate and optimize HPC workload managers, ensure high-throughput resource management, and remove kernel-space bottlenecks for massive parallel jobs.