I'm a researcher at CMU Catalyst Lab, working with Prof. Zhihao Jia. With Google CoreML, I'm building a compiler to lower Mirage-generated computation graphs into efficient Google TPU kernels. I'm also working on Mirage Persistent Kernel, the first compiler and runtime system that automatically transforms multi-GPU model inference into a single high-performance megakernel.

I worked on data systems with Prof. Vyas Sekar at the CyLab Security & Privacy Institute. I spoke at Current 2025 (a Confluent data conference), open-sourced FlinkSketch — probabilistic sketches for memory-efficient analytics on Apache Flink streams — and built ProjectASAP, low-latency data pipelines for agentic workloads.