Latest AI JobsBrowse Jobs
DigitalOcean logo

Principal Engineer, Model Optimizations

DigitalOceanSeattleAgentic AI
Apply on company site
Salary
$249,600 - $312K
Job type
Full-time
Experience
Senior · 12+ yrs
Posted
8 Oct 2026
Status
Open · checked 7d ago

About this role

Principal Engineer, Model Optimizations is a full-time hybrid role based in Seattle at DigitalOcean. This position suits an experienced performance systems engineer looking to lead model optimization strategies across a heterogeneous GPU fleet.

Role Overview

DigitalOcean is hiring a Principal Engineer, Model Optimizations for the Inference Platform team in Seattle. The person in this role will own the model optimization discipline end to end, covering quantization strategies, kernel selection, speculative decoding, and parallelism layouts across NVIDIA and AMD hardware.

Key Responsibilities

Setting the technical strategy for model optimization across the fleet, deciding which techniques to invest in or consume upstream.
Owning quantization end to end, including calibration methodology, accuracy budgets, and evaluation gates.
Driving performance for modern architectures at the execution level, such as MoE routing, attention variants, and memory-movement patterns.
Leading speculative decoding work and acceptance-rate tuning.
Writing and tuning kernels where upstream falls short across CUDA, Triton, CUTLASS, and ROCm counterparts.
Establishing repeatable parallelism layout strategies per model, GPU family, and traffic shape.
Building benchmarking and regression infrastructure for trustworthy optimization claims.
Making AMD a first-class target and driving upstream work in vLLM, SGLang, and TensorRT-LLM.
Partnering with NVIDIA and AMD engineering teams on pre-silicon enablement and roadmap feedback.
Setting technical direction, mentoring engineers, and representing DigitalOcean in upstream communities and conferences.

Requirements

12+ years in performance-critical systems with substantial recent experience optimizing LLM inference in production.
Deep understanding of GPU architecture and the inference performance model.
Hands-on kernel-level experience on at least one vendor stack and the ability to work across both CUDA and ROCm.
Practical quantization expertise with the judgment to evaluate customer accuracy bars.
Familiarity with the internals of at least one major serving engine like vLLM, SGLang, or TensorRT-LLM.
Strong Python, C++, and CUDA skills, plus profiling rigor with Nsight and rocprof.
A measurement-first disposition with reproducible benchmarks.
Excellent written and verbal communication skills with cross-functional leadership experience.

Nice to Have

Meaningful upstream contributions to vLLM, SGLang, TensorRT-LLM, PyTorch, Triton, or ROCm.
Experience bringing up a new accelerator family for production inference.
Experience with disaggregated prefill/decode serving.
Track record of day-zero enablement for newly released frontier open models.
Publications or patents in efficient inference, quantization, or GPU kernel design.

Skills and Tools

Python
C++
CUDA
Triton
CUTLASS
ROCm
HIP
Composable Kernel
hipBLASLt
AITER
vLLM
SGLang
TensorRT-LLM
PyTorch
Nsight
rocprof

Salary and Benefits

The salary range for this position is $249,600 to $312,000. Additional compensation elements may include a performance bonus and equity compensation such as hire grants and an Employee Stock Purchase Program. DigitalOcean also offers a competitive array of well-being benefits, flexible time off, employee assistance programs, local meetups, and career development resources including conference reimbursements, training support, and access to LinkedIn Learning courses.

About the Company

DigitalOcean provides an inference cloud with a strong community of top talent focused on building a simple, scalable cloud for builders and dreamers.

How to Apply

Apply on the official company careers page.

Skills and tech stack

PythonC++CUDATritonCUTLASSROCmHIPComposable KernelhipBLASLtAITERvLLMSGLangTensorRT-LLMPyTorchNsightrocprofLLM

Questions candidates ask

What is the location of the Principal Engineer, Model Optimizations role?

The role is a hybrid position located in Seattle.

What is the salary range for this position?

The salary range is between $249,600 and $312,000, with potential for bonuses and equity compensation.

How many years of experience are required?

Candidates must have 12+ years in performance-critical systems, with substantial recent experience optimizing LLM inference in production.

What hardware stacks does the role work with?

The role works across a heterogeneous GPU fleet including both NVIDIA and AMD hardware stacks.

Is visa sponsorship available?

Visa sponsorship details are not stated in the job posting.

Ready to apply?

The application is on the official DigitalOcean careers page.

Apply on company site