T Vidhya

I’m a Software Engineer specializing in Machine Learning Systems and Performance Optimization, with practical experience designing and tuning training and inference pipelines across AWS and Azure environments. My focus is on reducing latency, maximizing throughput, and optimizing compute efficiency for large-scale AI workloads. I work hands-on with PyTorch, TensorFlow, ONNX Runtime, and TensorRT, profiling model execution with Nsight Systems, VTune, and PyTorch Profiler to improve kernel performance and GPU utilization. Recently, I’ve optimized GenAI and LLM inference workloads by implementing mixed-precision (FP16/BF16) training, dynamic batching, and quantized deployments on AWS Trainium/Inferentia. My background in software engineering (Java + Python) and cloud-native microservices (Kubernetes, Terraform) allows me to bridge ML and systems engineering to deliver production-grade performance improvements. Areas of Interest: ML Inference Optimization • Kernel Fusion • GPU Profiling • Distributed Training • AI Infrastructure • AWS Silicon Ecosystem (Trainium, Inferentia)