work
| Google Software Engineer | 2025.11 - Present |
| Lead development of the PyTorch-native vLLM TPU backend and co-develop TorchTPU, focusing on serving performance, kernels, and quantization. | |
| AWS, Annapurna Labs Software Engineer | 2025.3 - 2025.11 |
| Founding NKI engineer, optimizing Trainium compiler passes, FlashAttention kernels, and autotuning. | |
| AWS, Annapurna Labs Software Engineer Intern | 2024.5 - 2024.8 |
| Built compiler autotuning and profiling infrastructure, improving Llama 3.1 performance by 14.7% and accelerating tuning workflows by 8.6x. | |
| TikTok Software Engineer Intern | 2022.5 - 2022.8 |
| Built low-latency avatar animation systems for AR/VR, including motion matching and skeleton retargeting. | |
| University of California, Los Angeles Research Assistant | 2022.6 - 2022.11 |
| Developed GNN cost models and ML/RL-driven design-space exploration for FPGA accelerators. | |