Sunil Reddy Aramreddys
Graduate Research Assistant @Umkc School Of Science And Engineering
Signup · Get unlimited contacts
WORK HISTORY
Graduate Research Assistant @Umkc School Of Science And Engineering
Kansas City, MO, US
Project: Robustness Analysis of AI Inference under RF/EMI Interference• Built a complete end-to-end AI inference pipeline on NVIDIA Jetson using ImageNet models (ResNet-18/50, MobileNet), including dataset preparation, deterministic preprocessing, and PyTorch → ONNX → • TensorRT (FP16/INT8) deployment.• Designed controlled evaluation workflows to measure latency, accuracy, and numerical stability under fixed power, clock, and runtime conditions.• Implemented large-scale bit-flip fault-injection experiments on both weights and activations (sign, exponent, mantissa) across multiple ResNet-18 layers to quantify robustness.• Developed fixed-input vs per-image evaluation modes to separate deterministic repeatability from input-dependent vulnerability.• Built an automated fault-injection pipeline with per-layer parameter sweeps, CSV logging, and aggregation scripts, enabling analysis over thousands of inferences.• Characterized gradual degradation and catastrophic failure regimes using multiple metrics (top-1 accuracy change, NaN/all-NaN rates, L1 and max-absolute output drift).• Designed a deterministic RF/EMI experiment timeline separating warm-up, synthetic workload, and DNN inference for clean RF alignment.• Implemented GPIO-based LED timing markers to synchronize AI inference phases with external RF measurement equipment.• Added a hardware-level anomaly indicator (second LED) using MSE-based classification-head monitoring to detect silent output drift beyond top-1 accuracy changes.• Built GPU-only, CPU-only, and mixed-workload baselines to isolate electromagnetic emissions from individual compute subsystems.• Collaborating with an interdisciplinary EMI lab to correlate RF measurements with AI inference behavior and reviewing Winograd convolution as a potential algorithm-level factor in EMI susceptibility.
EDUCATION
University of Missouri-Kansas City
Doctor of Philosophy - PhD, Computer Science
MADANAPALLE INSTITUTE OF TECHNOLOGY & SCIENCE
Bachelor's degree, Computer Science
University of Missouri-Kansas City
Master's degree, Computer Science
ABOUT SUNIL REDDY ARAMREDDYS
Graduate student passionate about turning AI and biometric innovations into real-world security solutions. Experienced in deep learning, computer vision, and full-stack deployment (Java, AWS, Docker). Focused on building secure and scalable technologies for the future of identity and trust.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.