NVIDIA Data AI Engineer

Hexaware Technologies is hiring an NVIDIA Data AI Engineer to deploy and operate NVIDIA NIM inference services and build production data pipelines for OCR and document understanding. The role focuses on GPU sizing and performance optimization, Kubernetes deployment and scaling, and reliable productionization of inference services.

Responsibilities:

• Deploy and operate NVIDIA NIM microservices (including OCR, vision, and LLM NIMs) on appropriately sized GPU configurations, selecting the right GPU type, count, and memory profile for each workload
• Right-size and tune GPU configs (GPU model, MIG partitioning, MPS, batch size, concurrency) to balance throughput, latency, and cost for production inference
• Build and orchestrate data engineering pipelines for OCR - document ingestion, preprocessing, layout/text extraction, and post-processing of extracted •content.•
• Integrate NIM inference endpoints into pipelines for OCR, document understanding, embeddings, and downstream inferencing
• Maximize GPU utilization and pipeline throughput using dynamic/continuous batching, async request handling, and parallel processing across documents/pages
• Deploy and scale NIM containers on Kubernetes using the NVIDIA GPU Operator, device plugins, and autoscaling for variable inference demand
• Monitor and profile GPU and pipeline performance (utilization, latency, queue depth, memory) using DCGM, NVidia Smi, and observability tooling; troubleshoot •bottlenecks.•
• Build robust, fault-tolerant data pipelines using orchestration tools (Airflow, ADF, or similar) with retries, checkpointing, and back-pressure handling
• Manage NIM model artifacts via the NGC catalog, configure model profiles, and apply optimized engines (TensorRT) shipped within NIM
• Collaborate with data and platform teams to productionize, version, and operate inference services with CI/CD

Requirements:

• Bachelor's/master's in computer science, Engineering, or related field
• 3+ years building production data engineering and/or AI inference systems
• PDF/image document processing: OCR, layout, tables, reading order, and metadata
• Evaluation using ground truth, confidence calibration, and human-defined validation rules
• Nemotron Parse 2.0 / NeMo Retriever OCR
• Strong proficiency in Python for pipeline and integration development
• Hands-on experience deploying GPU-accelerated inference services - ideally NVIDIA NIM
• Experience building data pipelines and processing large volumes of documents/files (OCR, document AI, or similar)
• Solid understanding of GPU configurations and how to size them for inference (memory, batch size, concurrency, MIG/MPS)
• Experience with containerization (Docker) and orchestration (Kubernetes), including GPU scheduling
• Comfort with parallel/concurrent processing for high-throughput pipelines
• Direct experience with NVIDIA NIM microservices and the NGC catalog
• Experience with OCR / document-intelligence engines and pipelines
• Familiarity with Triton, TensorRT, and serving optimization (batching, quantization profiles)
• GPU partitioning experience (MIG, MPS) for multi-tenant inference
• Experience with RAPIDS/DALI for GPU-accelerated data preprocessing