Senior GPU Systems Engineer - CUDA & PyTorch

Bright Vision Technologies is a technology consulting and software development firm delivering cloud, AI, data, and enterprise solutions across the United States.

We are seeking a GPU Software Engineer with deep expertise in CUDA C/C++ programming, modern GPU architectures, and high-performance computing (HPC) to optimize compute-intensive workloads across modern accelerator hardware. In this role, you will extract maximum throughput from GPU platforms for AI training, LLM inference, scientific computing, and high-throughput data pipelines. You will write custom CUDA kernels, optimize memory access patterns across HBM, L2, shared memory, and registers, and build multi-GPU distributed communication pipelines using NCCL and RDMA. Working closely with machine learning engineers and systems architects, you will debug performance bottlenecks, write Triton or C++ operators, and mentor engineering team members on low-level performance profiling.

Key Responsibilities

• High-Performance Kernel Development: Design, implement, and optimize custom CUDA C/C++ kernels for deep learning primitives (attention mechanisms, GEMM, linear algebra, custom activations).
• Profiling & Performance Tuning: Profile GPU workloads using NVIDIA Nsight Systems, Nsight Compute, and CUDA profilers to optimize thread occupancy, register pressure, and memory bandwidth utilization.
• Distributed & Multi-GPU Systems: Scale multi-GPU and multi-node training/inference pipelines utilizing NCCL, RDMA, and high-performance networking fabrics.
• ML Framework Integration: Implement custom operators and fused kernels in modern ML frameworks (PyTorch, JAX) and kernel-authoring environments (Triton, CUTLASS).
• Quantization & Mixed-Precision: Implement FP16, BF16, FP8, and quantized compute paths that maximize hardware tensor core throughput while preserving numerical stability.
• Benchmarking & Regression Testing: Establish automated benchmarking suites and regression tests to safeguard GPU kernel performance across hardware generations.
• Technical Mentorship & Governance: Lead code and design reviews, establish tuning playbooks, and mentor junior engineers on GPU architecture execution models.

Requirements

Systems Programming & CUDA Depth

• 10+ years of hands-on professional software engineering experience specializing in GPU programming, parallel computing, and systems performance engineering.
• CUDA & C++ Mastery: Deep expertise in CUDA C/C++ and modern C++ systems programming practices.
• GPU Architecture Expertise: Thorough understanding of modern GPU hardware architectures (NVIDIA Ampere, Hopper, Blackwell), thread execution models (warps, thread blocks), and memory hierarchies (HBM, L2, shared memory, registers).
• Profiling & Telemetry: Proven track record profiling and optimizing production GPU workloads using NVIDIA Nsight tools.
• Distributed Computing: Practical experience with multi-GPU interconnects and communication libraries (NCCL, MPI).
• Framework Integration: Demonstrated success writing and binding custom C++/CUDA operators into PyTorch, JAX, or C++ execution engines.
• Numerical Foundations: Solid grounding in linear algebra, parallel algorithms, and numerical methods.

Education & Work Logistics

• Bachelor’s or Master’s degree in Computer Science, Computer Engineering, or a related technical discipline.
• Ability to work 100% remotely within the United States.

Preferred Qualifications

• Experience authoring kernels using Triton or CUTLASS frameworks.
• Familiarity with LLM serving and inference engines (vLLM, TensorRT-LLM, FasterTransformer).
• Exposure to compiler infrastructure (LLVM, MLIR, or TVM) for tensor codegen.
• Active open-source contributions to GPU performance libraries or distributed AI projects.

Total Rewards & Benefits

• Compensation: Competitive base salary ($100,000 – $175,000 USD W2) + performance reviews.
• Workplace Flexibility: 100% remote work model across the United States.
• Growth & Impact: High-visibility technical leadership opportunity shaping AI acceleration infrastructure.
• Equal Opportunity Employer: BV Teck is an EEO employer committed to inclusive hiring across technical specialties.