Los Angeles, California, United States
Computer Engineering PhD student at USC working on computer architecture and energy-efficient AI systems. I focus on hardware/software co-design for ML accelerators, including compiler-guided power management, DVFS, memory hierarchy design, and accelerator scheduling under real-world power, area, and latency constraints. My experience spans performance modeling, RTL design, synthesis, physical design, and tapeout for AI accelerator systems.
Developed machine learning model monitoring and retraining loop on Athinia Optimized scalable cloud-based data science and analytics workflows to reduce computational costs and enhance efficiency Developed ML pipelines that streamlined data processing and improved model accuracy in a distributed setting on Foundry Supporting customer analytics applications with ML engineering expertise to increase customer satisfaction and retention
Authored a paper on hardware accelerators for heterogeneous platforms; Published in FPL 2023 Accelerated the inference process of various GNN models on the Xilinx Adaptive Compute Acceleration Platform (ACAP) VCK5000 board with processors, FPGA, and AI Engine. Streamlined different matrix computations based on their sparsity onto the AI Engine and FPGA, achieving 162.42× speedup Assessed and integrated GNN models utilizing PyTorch Geometry in Python and datasets such as Planetoid, Flicker, and Reddit
Attend weekly labs and office hours to help students debug Verilog and FPGA problems Graded and provided comments on students’ assignments and exams in C++
Designed a low-cost wearable induction-based sensor control circuit to recognize human motion Simulate the circuit via LTspice (simulator software) Layout Print Circuit Board circuit via KiCad EDA software
Developed a multi-node simulation for data collection using a hexacopter on ROS, collecting information to analyze structures Applied image processing and applied computer vision methods for collecting data for analysis Created synthetic crack generator with multiple different timestamps on an area, which accelerated the data collection process
Managed databases, server, and label systems, enhancing integration with packaging and testing teams Developed a digital request function in C#, phasing out paper forms and significantly reducing manual process time