Siddharth Samsi, Ph.D.
Siddharth Samsi, Ph.D. Siddharth Samsi is a Senior Solutions Architect and Data Scientist within the Financial Services and Technology team at NVIDIA. In this capacity, he specializes in GPU applications and accelerated computing initiatives for the financial services industry. His expertise encompasses high performance computing (HPC), distributed artificial intelligence (AI)training, time series forecasting, and performance optimization. Before joining NVIDIA, Dr. Samsi led research at the MIT Lincoln Laboratory Supercomputing Center. There, his work focused on distributed AI, high performance computing, and energy-aware computing strategies for artificial intelligence applications. Dr. Samsi earned both his Ph.D. and master’s degrees in electrical and computer engineering from The Ohio State University.
Abstract
Performance, Productivity, and Power: GPU-Accelerated Workflows from Network Analysis to Graph Neural Networks for Fraud Detection
AI workloads span from exploratory, single-GPU workflows to datacenter-scale deployments, demanding a balance of performance, productivity, and power efficiency. This talk will start with a brief walkthrough of the Anonymized Network Sensing Graph Challenge, re‑implementing GraphBLAS‑style formulations using commodity data‑science stacks to achieve over 400x speedups over CPU implementations, without custom HPC code. Next, we will discuss a case study for financial fraud detection where Graph Neural Networks and transformers are reshaping fraud/risk pipelines and research workflows. Finally, we will share recent work driving greater energy efficiency and the impact of developer choices on power, energy, and performance for HPC and AI applications running on NVIDIA technology.