Looking for a job

Narasimha Chari Nanchella

@narasimhacharinanche

AI/ML Engineer at NVIDIA building GPU-accelerated LLM, speech, and vision services, with inference latency reduced by approximately 3x.

United States

Message

At NVIDIA, I design and deploy GPU-accelerated AI microservices for production LLM, speech, and vision workloads. I optimized PyTorch and Hugging Face Transformer models, reducing inference latency by approximately 3x and GPU memory consumption by 45%.

At Microsoft, I developed computer vision and video analytics pipelines to identify driver behavior across video datasets. I also built distributed training workflows that improved training efficiency by approximately 35%, and optimized inference to achieve sub-100ms latency for real-time prediction services.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer

Current

NVIDIA

Aug 2025 - Present (1 year 3 months)

Designed and deployed GPU-accelerated AI microservices using Python, FastAPI, Triton Inference Server, TensorRT-LLM, CUDA, and REST/gRPC APIs. Optimized PyTorch and Hugging Face Transformer models using ONNX, TensorRT, FP16, and INT8 quantization, reducing inference latency by approximately 3x and GPU memory consumption by 45%.

AI/ML Engineer

Microsoft

Apr 2020 - Jun 2024 (4 years 2 months)

Developed computer vision and video analytics pipelines using Python, PyTorch, OpenCV, CNN architectures, and transfer learning to identify driver behavior. Built data processing pipelines using Python, Apache Spark, Airflow, Hive, and Presto to ingest and transform over 50,000 hours of video into structured datasets.