GPU Kernel Optimization Lab
Kernel EngineeringTiled matrix multiplication kernels evaluated under simulated 64 KB SRAM budget with HLS loop pipelining and array partitioning.
4.1× throughput improvement on 512×512 matrices
Swarnalata Panigrahy
ML systems, from graph to hardware.
Current research
CORSA Lab, UC Irvine
NeuSym-HLS: Partial symbolic distillation and high-level synthesis for compact, hardware-aware edge inference. Replaces selected DNN layers with analytic expressions through symbolic regression, evaluating across accuracy, latency, and hardware resources.
Selected work
Inference Serving System
Inference SystemsProduction recommendation engine with batched prediction, result caching, and optimized output schemas across international markets.
31% latency improvement
Mini Tensor Compiler
Compiler ConstructionHigh-level tensor expressions lowered into generated C++ loop kernels with shape tracking, validation, and fusion support.
60+ tensor shapes validated against PyTorch
NeuSym-HLS Research
ResearchPartial symbolic distillation and high-level synthesis for compact edge inference models with hardware-aware evaluation.
Active research at CORSA Lab
Kria KR260 Accelerator
Accelerator SystemsFPGA offload inference stack with PYNQ host control, programmable logic, and accuracy validation within 1% of software baseline.
3.2× speedup over ARM CPU baseline
Capabilities
Inference systems
Batch optimization, result caching, output schema design
Kernel engineering
Tiled kernels, HLS pipelining, memory hierarchy optimization
Compiler construction
IR lowering, code generation, shape tracking, operator fusion
Accelerator systems
FPGA offload, PYNQ, hardware/software co-design
Research & Experience
2025—present
Graduate Researcher
CORSA Lab, UC Irvine
NeuSym-HLS: Symbolic regression for DNN compression with hardware-aware synthesis.
2022—2024
Associate Engineer — Machine Learning
Ingram Micro
Built production recommendation engine with 90% catalog coverage across 28 countries. Validated ranking quality and data consistency internationally.
2020—2022
Graduate Studies
University of California, Irvine
Master's in Electrical Engineering and Computer Science. Focus on machine learning systems and hardware acceleration.