Staff Platform Engineer - GPU Inference Control Plane

Fdetimes; Greater London; On-site; GBP 90,000 - 120,000; Full time

Job summary

Together AI, the AI Native Cloud, seeks a software engineer to build a Kubernetes-native control plane that provisions and runs a GPU inference fleet. You’ll design a manifest-driven API for clusters, models, and capacity changes, with controllers handling reconciliation and lifecycle management behind a single API.

Expect high emphasis on fragmentation, bin-packing, and efficient scheduling. You will own reliability, CI/CD for infrastructure code, and a self-service surface (API/CLI) while

Qualifications

• Strong software engineering background in Go, Python, Rust, or similar.
• Experience with durable workflow orchestration tools such as Temporal, Cadence, or equivalent.
• Experience building control planes or orchestration systems (e.g., Kubernetes controllers/operators).
• Experience with event-driven systems (Kafka, NATS, SQS).
• A product mindset with internal platforms or APIs consumed by other teams.

Responsibilities

• Build the provisioning state machine modeling full lifecycle of a host from discovery to decommission.
• Create a self-service API for declaring and scaling inference clusters via one API call.
• Automate self-healing: drain, repair, and reintroduce capacity automatically.
• Ensure reliability: idempotency, retries, rollback, drift detection.
• Collaborate with ML platform to encode cluster shapes as abstractions in the platform.
• Write software with strong typing, tests, code review, versioning, and CI/CD.

Skills

Go

Python

Rust

Temporal

Cadence

Kubernetes controllers

event-driven systems

Kafka/NATS/SQS

product mindset

Tools

Kubernetes

Temporal

Cadence

Kafka/NATS/SQS

Job description

Together AI, the AI Native Cloud, seeks a software engineer to build a Kubernetes-native control plane that provisions and runs a GPU inference fleet. You’ll design a manifest-driven API for clusters, models, and capacity changes, with controllers handling reconciliation and lifecycle management behind a single API.

Expect high emphasis on fragmentation, bin-packing, and efficient scheduling. You will own reliability, CI/CD for infrastructure code, and a self-service surface (API/CLI) while