Gramian Consultancy is a boutique consultancy specializing in IT professional services and engineering talent solutions. With a strong background in software engineering and leadership, we help companies build high-performing teams by matching them with professionals who truly fit their needs.Role OverviewWe are looking for experienced CUDA Developers to work on advanced AI and machine learning initiatives focused on improving the capabilities of large language models (LLMs). In this role, you will solve complex GPU programming challenges, optimize high-performance CUDA workloads, review AI-generated code, and contribute to the development of more capable AI systems.Duration: 3 monthsCommitment: 40h/week, 4h/day overlap with PSTModel: Contract, time and materialLocation: 100% Remote: Bangladesh, Brazil, Colombia, Egypt, Ghana, India, Pakistan, Indonesia, Kenya, Nigeria, Turkey, VietnamInterview: 1 technical interviewKey ResponsibilitiesSolve advanced CUDA and GPU programming problems involving parallel computing and performance optimization Review, evaluate, and improve AI-generated CUDA, C++, and Python code Optimize GPU kernels for throughput, latency, memory efficiency, and resource utilization Work with CUDA libraries and frameworks such as Thrust, cuBLAS, and cuDNN Debug and resolve issues related to CUDA kernels, synchronization, and memory management Develop high-quality technical prompts, solutions, explanations, and evaluations for AI model training Collaborate with AI researchers, engineers, and evaluation teams Stay up to date with the latest developments in CUDA, GPU architectures, and performance optimization techniquesRequirements5+ years of professional software development experience with strong focus on CUDA development Strong proficiency in C/C++ Strong hands-on experience with Python and scientific computing ecosystems Experience working with PyTorch and NumPy Experience with CUDA 12.3 or newer Strong understanding of GPU programming, parallel computing, and performance optimization Experience optimizing workloads for high-performance execution and efficient resource utilization Experience with CUDA libraries such as Thrust, cuBLAS, and cuDNN