Remote Staff Engineer, AI Storage & GPU Infra

Together Computer Inc; San Francisco (CA); Ibrido; USD 250.000 - 330.000; Tempo pieno

Vantaggi offerti da questo lavoro

Remote work flexibility

Equity

Health insurance

Descrizione del lavoro

Together AI seeks an accomplished storage engineer to lead and scale multi-petabyte AI/ML storage systems. You will integrate Vast, Weka, Ceph, and Lustre, and build Kubernetes-native storage operators to automate provisioning and enforce strict multi-tenancy at cluster scale.

You will optimize data paths for 10+ GB/s per node, implement intelligent caching for model weights and KV-cache offload, and contribute to open-source storage tooling while collaborating with inference engines and

Competenze

• 8+ years in storage engineering at multi-petabyte scale.
• Proven track record with GPU/HPC storage for training/inference.
• Experience with KV-cache offload or disaggregated KV storage.
• Deep Kubernetes and cloud-native storage in production.
• Strong coding in Go and Python for production systems.

Mansioni

• Define storage strategy and roadmap for Together AI.
• Scale multi-petabyte AI/ML storage with Vast, Weka, Ceph.
• Develop caching and tiered storage for extreme IOPS at GPU scale.
• Design KV-cache offload and reuse across GPU memory and NVMe layers.
• Ensure multi-tenant isolation at scale; build storage operators.

Conoscenze

Go

Python

C/C++

Rust

Kubernetes

Terraform

GitOps

Distributed storage

Formazione

BS/MS in CS or related

Strumenti

Ceph

WekaFS

Lustre

Vast

GPFS

S3

MinIO

TensorRT-LLM

Descrizione del lavoro

Together AI seeks an accomplished storage engineer to lead and scale multi-petabyte AI/ML storage systems. You will integrate Vast, Weka, Ceph, and Lustre, and build Kubernetes-native storage operators to automate provisioning and enforce strict multi-tenancy at cluster scale.

You will optimize data paths for 10+ GB/s per node, implement intelligent caching for model weights and KV-cache offload, and contribute to open-source storage tooling while collaborating with inference engines and