About the job

We build production AI agent systems for enterprise clients across financial services, health, government and telco, where “it worked in the demo” isn’t good enough. This isn’t prompt engineering or wiring an LLM to a vector database. You’ll be doing harness design, loop engineering, context engineering and spec-driven builds: the actual discipline behind agents that hold up in production, under audit, at a client that can’t tolerate a shadow AI incident.

This is also DX1’s feeder role into our Forward-Deployed Engineer programme. Engineers who build well here, meet the FDE selection criteria and complete the certification pathway are on track for conditional promotion into FDE, where you embed directly with clients and own delivery end to end.

Key Responsibilities

• Design and build production-grade AI systems: agentic architectures, multi-model orchestration, RAG pipelines, tool use and MCP integrations.
• Build and run evaluation frameworks so agent behaviour in production is measurable and defensible: eval suites, regression tests, OpenTelemetry-based tracing and monitoring.
• Design the harness an agent operates inside (tools, permissions, state, guardrails) and the loop it runs (plan, act, observe, correct), not just the prompt in front of it.
• Work from clear specs that hold up under review, taking systems from prototype through to deployment, CI/CD and operational runbooks.
• Apply AI-specific security controls: input validation, output controls, PII detection, audit trails, and protections against prompt injection and excessive agency.
• Contribute reusable assets (agent templates, integration patterns, compliance packs) to DX1’s asset library so every engagement after yours starts further ahead.
• Required Qualifications and Experience
• 2+ years in production software engineering, with systems you’ve actually shipped and operated.
• Production experience building on LLMs: RAG pipelines, agent development, prompt and context engineering, evaluation.
• Strong Python and/or TypeScript.
• Working knowledge of cloud platforms (AWS preferred) and containerised deployment.
• Experience instrumenting production systems with OpenTelemetry for tracing and observability, ideally extended to agent or LLM workloads (tool calls, reasoning steps, latency and cost per step).
• Ability to work from a scoped spec and deliver without close supervision, and to explain technical tradeoffs to non-technical stakeholders.

Certification and Growth Path

DX1 funds your certification path from day one: AWS Certified AI Practitioner and Anthropic’s Claude Certified Architect (CCA) Foundations within your first six months. These double as the two pre-requisites for progression into the FDE programme. DX1 also holds AWS partner accreditation for generative AI.

Bonus Points

• Experience deploying agents on AWS Bedrock with Anthropic models (or a similar managed model platform), rather than just calling a raw API.
• Background in consulting or client-facing delivery.
• Knowledge of Australian data privacy regulations and AI governance frameworks.