CUDA Engineer

Overview

As a CUDA Engineer at Fuse Energy, you will design and optimize low-level GPU kernels that power transformer inference workloads in a fast-growing renewable energy and AI compute environment. You will push throughput and latency, tune memory hierarchies, and implement quantisation-aware, mixed-precision kernels across cutting-edge GPUs. Your work directly supports energy-efficient AI inference at scale, helping Fuse meet rapid compute demands while aligning with the company’s mission to accelerate clean energy via advanced technology. You will collaborate with cross-functional teams to improve performance, reliability, and efficiency of our inference pipelines.

Pay / Benefits

• Competitive salary and an equity sign-on bonus
• Biannual bonus scheme
• Fully expensed tech to match your needs
• Breakfast and dinner allowance for office based employees

Responsibilities

• Write and optimize custom CUDA kernels for core transformer inference operations
• Profile kernels to identify bottlenecks in occupancy, memory throughput, and warp divergence
• Apply kernel fusion to reduce memory round-trips and launch overhead across inference pipelines
• Optimize memory access patterns and manage the memory hierarchy for maximum bandwidth utilization
• Implement quantisation-aware kernels and mixed-precision arithmetic to reduce latency and memory footprint
• Build and tune caching mechanisms for efficient autoregressive decoding
• Tune kernel launch configurations for target GPU architectures
• Benchmark kernels against baselines and drive throughput/latency improvements
• Write tests for CUDA code to catch performance and correctness regressions
• Maintain internal CUDA libraries and contribute to team coding standards and documentation

Key requirements

• 4+ years writing production CUDA code with a track record of shipping performance-critical kernels
• Deep understanding of GPU microarchitecture, warps, occupancy, register pressure, and memory hierarchy
• Strong CUDA C++ skills, including streams and asynchronous execution
• Hands-on experience profiling to diagnose compute-bound vs memory-bound bottlenecks
• Experience with kernel fusion, memory coalescing, and avoiding warp divergence
• Experience writing quantised and mixed-precision kernels
• Solid grasp of parallel algorithm design and numerical precision tradeoffs

• CUDA C++
• GPU profiling
• kernel fusion
• memory hierarchy optimization
• quantisation-aware kernels
• mixed-precision arithmetic

…