Principal Engineer, Model Optimizations

DigitalOcean

. Set the technical strategy for model optimization across DigitalOcean's heterogeneous NVIDIA and AMD accelerator fleet .

Posted 10/8/2026full-timeSeattle • Washington • United StatesLead💰 $249,600 - $312,000 per yearWebsite

Core Competencies

Role fit

Core Competencies

Use this summary to align your resume positioning with the role.

Demonstrates extensive expertise in optimizing LLM inference performance across diverse GPU architectures, with a strong focus on quantization strategies and kernel-level programming. Proven ability to lead technical direction, mentor engineers, and collaborate with cross-functional teams to drive impactful results.

Highest-signal resume keywords

LLM Inference OptimizationCUDA ProgrammingQuantization ExpertiseGPU Architecture UnderstandingCross-Functional Leadership

ATS Keywords

Tailor your resume

Applicant Tracking System Keywords

Tip: use these terms in your resume and cover letter to boost ATS matches.

Hard Skills

CUDAC++PythonKernel-Level ProgrammingQuantizationPerformance OptimizationBenchmarkingProfilingModel EvaluationTraffic Distribution Analysis

Soft Skills

Excellent CommunicationMentoring

Tools & Technologies

NsightRocprofTritonCUTLASSHIPVLLMSGLangTensorRT-LLMAITERComposable Kernel

Industry Keywords

Performance-Critical SystemsInference PerformanceGPU FamiliesModel OptimizationTechnical Strategy

Tech Stack

Tools & technologies

PythonC++

About the role

Key responsibilities & impact

Set the technical strategy for model optimization across DigitalOcean's heterogeneous NVIDIA and AMD accelerator fleet
Own quantization strategy, calibration methodology, accuracy budgets, and evaluation gates
Drive execution-level performance for MoE, MLA, GQA, long-context, and sliding-window attention architectures
Lead speculative decoding work and acceptance-rate tuning
Write and tune CUDA, Triton, CUTLASS, HIP, Composable Kernel, hipBLASLt, and AITER kernels where upstream falls short
Establish repeatable parallelism layouts across models, GPU families, and traffic shapes
Build benchmarking and regression infrastructure for traffic distributions, TTFT, ITL, throughput, and accuracy validation
Make AMD a first-class target and drive upstream work in vLLM, SGLang, and TensorRT-LLM
Partner with NVIDIA and AMD engineering teams on pre-silicon enablement, early-access hardware, and roadmap feedback
Set technical direction, mentor senior and staff engineers, and represent DigitalOcean in upstream communities, conferences, and customer technical deep dives

Requirements

What you’ll need

12+ years in performance-critical systems, with substantial recent experience optimizing LLM inference in production
Deep understanding of GPU architecture and inference performance: memory bandwidth versus compute bounds, arithmetic intensity, kernel launch and scheduling overhead, and prefill versus decode differences
Hands-on kernel-level experience on at least one vendor stack and ability to work across CUDA/CUTLASS/Triton and ROCm/HIP/CK
Practical quantization expertise
Familiarity with vLLM, SGLang, or TensorRT-LLM internals sufficient to land non-trivial upstream changes
Strong Python and C++/CUDA skills
Profiling experience with Nsight, rocprof, or equivalent tooling
Measurement-first approach using reproducible benchmarks
Excellent written and verbal communication
Experience leading cross-functional efforts spanning infrastructure, product, and customers

Benefits

Comp & perks

Reimbursement for relevant conferences, training, and education
Access to LinkedIn Learning's 10,000+ courses
Employee Assistance Program
Local Employee Meetups
Flexible time off policy
Bonus eligibility based on company and individual performance
Equity compensation, including equity grants upon hire
Employee Stock Purchase Program
Hybrid work arrangement