Project Overview

A leading energy-sector company is looking for an experienced GPU Software Engineer with strong expertise in CUDA , GPU-accelerated numerical computation , and matrix operations .

The project focuses less on LLM/AI topics and instead centers on power flow calculations and large-scale numerical simulations that must be efficiently executed on NVIDIA GPUs .
You will work on porting, optimizing, and accelerating computational code onto CUDA , leveraging frameworks such as cuBLAS, cuSOLVER, cuSPARSE , or similar, as well as NVIDIA tooling (incl. QDSS , Jetson toolchain if relevant).

Key Responsibilities

• Porting and optimizing power flow / power system calculations to run on NVIDIA GPU hardware

• Designing and implementing high-performance CUDA kernels for matrix operations and numerical solvers

• Profiling and optimizing GPU execution using NVIDIA tooling (e.g., qdss , Nsight Systems/Compute)

• Working with large-scale matrix algebra , linear equation solving, iterative solvers, and sparse/dense matrix handling

• Adapting existing CPU-based simulation code to GPU environments

• Ensuring numerical stability and precision in GPU-accelerated computation

• Close collaboration with power system engineers and simulation experts

• Documentation and handover of GPU-optimized modules

• Optional: contribution to Jetson-based environments if needed

Required Skills

• Strong experience in CUDA development (custom kernels, memory management, warp optimization)

• Background in numerical linear algebra , matrix operations, and solving systems of equations

• Experience with GPU-accelerated libraries such as:

• cuBLAS, cuSOLVER, cuSPARSE, Thrust , or similar

Nice to Have

• Experience with power flow calculations , electrical grid simulation, or energy modeling

• Experience with NVIDIA Jetson platforms

• Familiarity with Python bindings (Numba/cuPy) or C++ integration

• Background in energy sector or critical infrastructure

• Knowledge of GPU cluster environments

Project Conditions

• Start: Flexible, ideally soon

• Duration: 6+ months with likely extension

• Mode: Remote / nearshore-friendly

• Onsite: Not required regularly

• Language: English