Tae Hoon Kim Email & Phone Number

San Diego, Ca, Us

Seattle, Washington, United States

▪ Integrated bottleneck analysis Python tools into a Golang and JavaScript-based profiler for custom machine learning hardware accelerators, streamlining performance analysis by eliminating the need for separate scripts, thus reducing analysis setup steps by 50%.▪ Improved profiler by reporting engine idle times (e.g., dependency waits) similar to Nvidia's Nsight-compute profiler, enabling users to seamlessly view bottleneck analysis alongside existing data such as engine instructions

Pittsburgh, Pennsylvania, United States

Conducted research within the Computer Architecture and Operating Systems Group (CAOS)▪ Published “Perspective: A Principled Framework for Pliable and Secure Speculation in Operating Systems” as the First Author to ISCA 2024 under mentorship of Professor Dimitrios Skarlatos▪ Divided the monolithic linux kernel address space into instruction and data "speculation views" that only allow speculative access or execution within the speculation views to improve performance while guaranteeing security▪ Compared to state-of-the-art defense, reduced OS system call latency overhead from 47.5% to 3.5% and lowered datacenter application throughput overhead (httpd, nginx, redis, memcached) from 5.7% to 1.2%, measured in gem5, a detailed CPU simulator in C++.

Master Of Science - Ms, Computer Science - Research Thesis, 3.89

Bachelor Of Science - Bs, Computer Science, 3.97

Minor in Engineering Studies, Minor in Business Administration