Machine Learning Engineer

Acceler8 Talent; San Francisco (CA); Ibrido; USD 160.000 - 260.000; Tempo pieno

Descrizione del lavoro

Acceler8 Talent is seeking a Machine Learning Engineer to optimize LLM inference performance in production. The role focuses on squeezing more performance from GPUs, distributed systems, and multi-node setups with Kubernetes-backed infrastructure.

You will work on high-performance infrastructure, profiling bottlenecks, and tuning batching, quantisation, and caching to scale AI workloads efficiently.

Competenze

• You'll come from an ML systems, distributed computing, HPC or performance engineering background and be comfortable working close to the hardware.
• Strong Python and/or C++, PyTorch and GPU computing experience are key.
• Experience with CUDA, NCCL, Triton, vLLM, TensorRT-LLM or SGLang would be particularly relevant.
• You should be comfortable reasoning about how GPU utilisation, memory bandwidth, communication overhead and model architecture interact to impact inference performance.

Mansioni

• Architect high-performance infrastructure for LLM inference
• Optimise latency, throughput and cost across production workloads
• Scale inference across multi-GPU and multi-node environments
• Improve GPU scheduling, allocation and utilisation
• Profile bottlenecks across compute, memory and networking
• Optimise batching, quantisation, KV caching and parallelism
• Scale GPU workloads across Kubernetes infrastructure

Conoscenze

Python

C++

PyTorch

CUDA

NCCL

Triton

vLLM

TensorRT-LLM

SGLang

GPU

Strumenti

Kubernetes

Descrizione del lavoro

Machine Learning Engineer

Want to work on the systems that make LLMs fast, scalable and efficient in production?

A Stanford-spun AI company is hiring anAI Inference Engineerto help build and optimise the infrastructure powering its production AI platform.

This is a deeply technical role focused on squeezing more performance fromGPUs, distributed systems and LLM inferenceat scale.

What you’ll be working on

• Architect high-performance infrastructure forLLM inference
• Optimiselatency, throughput and costacross production workloads
• Scale inference acrossmulti-GPU and multi-node environments
• Improve GPU scheduling, allocation and utilisation
• Profile bottlenecks acrosscompute, memory and networking
• Optimisebatching, quantisation, KV caching and parallelism
• Scale GPU workloads across Kubernetes infrastructure

What we’re looking for

You’ll come from anML systems, distributed computing, HPC or performance engineeringbackground and be comfortable working close to the hardware.

StrongPython and/or C++, PyTorch and GPU computing experience are key.

Experience withCUDA, NCCL, Triton, vLLM, TensorRT-LLM or SGLangwould be particularly relevant.

You should be comfortable reasoning about howGPU utilisation, memory bandwidth, communication overhead and model architectureinteract to impact inference performance.