LLM Inference Systems Engineer (Multi-Node, GPU)

ScOp Venture Capital LLC.; San Jose (CA); On-site; USD 150,000 - 350,000; Full time

Benefits offered by this job

Unlimited PTO

Full benefits

Two engineering-centric offices with

Job summary

ChipAgents is seeking an ML Systems Engineer to optimize the performance and efficiency of large language model inference powering our agentic AI platform. This role focuses on low-level systems optimization across multi-node clusters to push throughput and reduce latency in production workloads.

You will design, implement, and benchmark inference optimizations, profile bottlenecks from GPU kernels to memory, and collaborate with researchers to deploy new model architectures—rewarded with

Qualifications

• Advanced degree in CS/EE or equivalent experience.
• Experience with large-scale ML systems and high-performance inference.
• Proficiency in Python and C++/CUDA; familiarity with SGLang, vLLM, PyTorch.
• Strong systems debugging and profiling across GPU kernels to application logic.

Responsibilities

• Design, deploy, and optimize LLM inference systems across multi-node clusters.
• Implement and benchmark concrete inference optimizations.
• Profile and analyze inference bottlenecks from GPU kernels to memory bandwidth.
• Build evaluation harnesses and benchmarking frameworks measuring accuracy, throughput, latency, and resource usage.
• Collaborate with researchers to integrate new model architectures and optimizations into production inference infrastructure.
• Investigate and apply emerging techniques to continuously improve inference performance.

Job description

ChipAgents is seeking an ML Systems Engineer to optimize the performance and efficiency of large language model inference powering our agentic AI platform. This role focuses on low-level systems optimization across multi-node clusters to push throughput and reduce latency in production workloads.

You will design, implement, and benchmark inference optimizations, profile bottlenecks from GPU kernels to memory, and collaborate with researchers to deploy new model architectures—rewarded with