AI Platform Engineer

Build the platform that turns AI agents into reliable production systems.

You’ll own the AI platform layer that enables agents to complete complex, end-to-end work for enterprise customers. Your scope will cover agent runtime, context management, tool execution and orchestration, evaluation pipelines, system-level prompting, and observability.

This is a hands-on applied AI engineering role. You’ll build and operate systems powered by LLMs, retrieval, and agent orchestration, with direct accountability for their reliability, quality, latency, and cost.

You’ll also shape how models are used across the platform: choosing the right model for each task, routing work to more cost-effective models where appropriate, and deploying self-hosted open-source models on dedicated GPUs.

You’ll own sandbox behaviour and agent runtime logic, working closely with SRE, who own sandbox infrastructure and capacity.

As the product organisation expands into multiple initiatives, you’ll work alongside initiative leads, engineers, design engineers, and AI-focused engineers to turn platform capabilities into measurable customer impact.

• Build and evolve the agent runtime. Develop the context management, tool execution, and orchestration capabilities that agents need to work reliably in production.

• Make quality measurable. Build evaluation pipelines and harnesses that expose failure modes, catch regressions, and support confident releases.

• Improve reliability and performance. Diagnose failures across model behaviour, retrieval and data pipelines, tool calls, and infrastructure.

• Optimise model usage and cost. Evaluate trade-offs between capability, latency, and cost across hosted and self-hosted models.

• Make agent behaviour observable. Develop tracing and diagnostic tooling that helps the team understand what agents did, why they failed, and how to improve them.

• Own delivery end to end. Take ambiguous problems from design through implementation, evaluation, deployment, monitoring, and iteration.

We’re looking for practical experience and evidence of impact in the following areas:

• Production AI systems. You’ve shipped and operated LLM-powered systems beyond prototypes or research, with hands-on experience in prompting, context management, tool use, or agent orchestration.

• End-to-end ownership. You’ve turned ambiguous AI problems into production capabilities with measurable outcomes and remained accountable after launch.

• AI systems design. You design for non-deterministic outputs, context limits, and unreliable tool calls, treating reliability, cost, and latency as core requirements.

• Evaluation design. You can build meaningful test sets, define failure taxonomies, and create evaluations that reveal quality issues and regressions.

• Structured debugging. You investigate problems through clear hypotheses and trace failures across system boundaries to identify root causes.

• Technical judgement. You make sound build-versus-integrate decisions with incomplete evidence and adapt as models, providers, and tooling evolve.

• Designing and operating distributed systems with high throughput and complex failure modes.

• Securing AI systems against risks such as prompt injection, insecure output handling, and compromised dependencies.

• Contributing to open-source AI tooling or infrastructure.

This role centres on the systems that power AI agents. It does not involve training models or conducting ML research. Its primary focus is platform engineering rather than user-facing features or UI development.

• Languages: TypeScript and Python

• Cloud: GCP — Cloud Run, GKE, and GCS

• Infrastructure: Terraform and Docker

• Observability: Datadog

• Data: Postgres and Redis

• AI: Anthropic models and tooling, alongside self-hosted open-source models