High-Performance ML Inference Engineer

Reactor; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

Competitive SF salary

Early equity

Visa sponsorship

Relocation support

Health, dental, and vision coverage

Job summary

Reactor in San Francisco is seeking an ML Inference Engineer to maximize performance of generative media models and push ultra-low-latency, high-throughput inference. You will craft an in-house runtime, implement optimizations with PyTorch tools, and collaborate with partner teams to integrate external models.

The role requires deep expertise in PyTorch, TensorRT, CUDA, and model optimization techniques, with a focus on delivering cutting-edge inference capabilities at scale.

Qualifications

• Bachelor's degree in Computer Science, Electrical Engineering, or equivalent practical experience.
• Strong foundation in systems programming and bottleneck resolution.
• Expertise in PyTorch, TensorRT, TransformerEngine, Nsight, ONNX Runtime.

Responsibilities

• Drive model performance for diffusion models and high-throughput inference.
• Design and implement a high-performance in-house inference runtime.
• Apply optimizations using torch.compile, CUDA kernels, and specialized inference frameworks.
• Quantize, prune, and modify architectures to optimize neural networks.
• Profile and benchmark model performance to identify bottlenecks.
• Collaborate with model partner teams to integrate models into the platform.

Skills

PyTorch

TensorRT

CUDA

Model optimization

Quantization

Low-latency inference

GPU programming

Transformer architectures

Education

Bachelor's degree in CS/EE

Tools

Nsight

ONNX Runtime

TransformerEngine

Job description

Reactor in San Francisco is seeking an ML Inference Engineer to maximize performance of generative media models and push ultra-low-latency, high-throughput inference. You will craft an in-house runtime, implement optimizations with PyTorch tools, and collaborate with partner teams to integrate external models.

The role requires deep expertise in PyTorch, TensorRT, CUDA, and model optimization techniques, with a focus on delivering cutting-edge inference capabilities at scale.