Senior Engineer II, GPU Kernel and Performance

<div class="content-intro"><p>Dive in and do the best work of your career at DigitalOcean. Journey alongside a strong community of top talent who are relentless in their drive to build the simplest scalable cloud. If you have a growth mindset, naturally like to think big and bold, and are energized by the fast-paced environment of a true industry disruptor, you’ll find your place here. &nbsp;We value winning together—while learning, having fun, and making a profound difference for the dreamers and builders in the world.&nbsp;</p></div><p>DigitalOcean is seeking a <strong>Senior Engineer 2</strong> to play a key technical role in our GPU Kernel and Performance&nbsp;team. DigitalOcean aims to be the Inference Cloud of choice for digitally native companies and you will help ensure we can offer the industry-leading performance for our inference services. You will be responsible for the architectural decisions that maximize throughput and minimize latency for the world’s most advanced large models. As an IC leader, you will act as a force multiplier for the engineering organization, solving the most complex bottlenecks in memory bandwidth and compute utilization while guiding the technical roadmap for our high-performance inference fleet.</p> <h2><strong>What You’ll Do:</strong></h2> <ul> <li><strong>Kernel Engineering:</strong> Design and implement high-performance GPU kernels using <strong>Triton</strong> and <strong>CUDA C++</strong>.</li> <li><strong>Precision Optimization:</strong> Develop and deploy state-of-the-art quantization techniques (FP8, INT8, and experimental <strong>FP4</strong>) to double throughput without losing accuracy.</li> <li><strong>Memory Magic:</strong> Optimize memory access patterns (SRAM vs. HBM3e) to eliminate bottlenecks in long-context attention mechanisms.</li> <li><strong>Research to Production:</strong> Implement the latest architectural breakthroughs like <strong>FlashAttention-4</strong> and <strong>TileLang</strong> directly into our production stack.</li> </ul> <h2><strong>What You’ll Bring:</strong></h2> <ul> <li><strong>Low-Level Mastery:</strong> Deep understanding of GPU architectures (SMs, Warp scheduling, Tensor Cores).</li> <li><strong>The Toolkit:</strong> Expert-level Triton or CUDA. If you’ve contributed to the Triton compiler or wrote custom CUDA kernels for a major LLM, we want you.</li> <li><strong>Math Foundation:</strong> Strong grasp of linear algebra and how it maps to parallel hardware.</li> <li><strong>Performance Obsession:</strong> A track record of optimizing kernels to achieve &gt;80% of theoretical hardware pe