AI/ML Engineer at NVIDIA

Developed and deployed GPU-accelerated ML models (NLP and CV) using PyTorch Lightning, TensorFlow, and TensorRT, reducing inference latency by 35% in fraud detection and autonomous systems. Fine-tuned transformer models (BERT, GPT, mBERT, XLM-R) for multilingual NLP tasks, boosting accuracy by ~20%. Built real-time fraud detection pipelines with Kafka + Spark ML, incorporating anomaly detection and graph ML for risk scoring across 10M+ transactions daily. Collaborated with IT data engineering to design high-throughput pipelines on Snowflake, AWS, and Spark, ensuring ML datasets are reliably available. Trained CNN and YOLO-based object detectors for autonomous vehicles, optimized for GPU edge devices, reducing inference time by 40%. Integrated LLMs with customer data lakes via RAG, enabling contextual insights for segmentation and cross-sell analytics. Implemented MLOps using MLflow, Airflow, Docker, and Kubernetes for scalable deployment.