Senior Cloud Platform Engineer - GPU Inference Orchestration

Neura Market; Hinoba-an; On-site; PHP 2,322,000 - 3,980,000; Full time

Job summary

Together AI is seeking a software engineer to build a Kubernetes-native control plane that provisions and runs our GPU inference fleet. You will design a manifest-driven API where the inference team declares needs, while controllers handle reconciliation, provider/runtime selection, and lifecycle management under the hood.

You will also build systems for efficiency, including defragmentation and rebalancing, and scheduling/bin-packing improvements to maximize GPU utilization without sacrificing

Qualifications

• Strong software engineering background with modern languages (Go/Python/Rust) and production-grade software practices.
• Experience implementing durable workflows with tools like Temporal or Cadence and building control planes.
• Hands-on with Kubernetes controllers/operators or similar reconciliation loops to model state over time.
• Experience designing event-driven systems using message queues or streams (Kafka, NATS, SQS).
• A product mindset; building internal platforms or APIs used by other teams.

Responsibilities

• Build the provisioning state machine for full lifecycle management of hardware and GPU drivers.
• Create a self-service control plane API for one-click provisioning, scaling, and teardown of clusters.
• Automate self-healing: detect, drain, repair, and reintroduce healthy capacity automatically.
• Ensure reliability: idempotency, retries, rollback, drift detection in the deployment pipeline.
• Collaborate with the ML/inference platform team to encode topology and scheduling requirements.
• Treat infrastructure code as software: strong typing, tests, reviews, CI/CD.

Skills

Go

Python

Rust

Temporal

Kubernetes

Event-driven

Job description

Together AI is seeking a software engineer to build a Kubernetes-native control plane that provisions and runs our GPU inference fleet. You will design a manifest-driven API where the inference team declares needs, while controllers handle reconciliation, provider/runtime selection, and lifecycle management under the hood.

You will also build systems for efficiency, including defragmentation and rebalancing, and scheduling/bin-packing improvements to maximize GPU utilization without sacrificing