AI Platform Engineer
Fifty Mission Labs | Dublin, Ireland (hybrid) | Full-time, permanent
About us
Fifty Mission Labs builds world-class data and AI organisations. We work alongside our clients — from strategy through to production systems — on data transformation, generative AI adoption and technology modernisation. Small team, senior people, real delivery.
The role
We are hiring an AI Platform Engineer to build and run the platform layer our agentic systems sit on. This is a hands-on engineering role, not a research role. You will design the infrastructure, runtimes and guardrails that let agents do useful work reliably, securely and at a cost we can defend to a CFO.
You will work across multiple client engagements, so you need to be comfortable moving between codebases, cloud accounts and problem domains — and comfortable explaining your decisions to people who are not engineers.
What you'll do
- Design and build agent runtime infrastructure on AWS — Bedrock, AgentCore, and the surrounding services (Lambda, ECS, Step Functions, API Gateway, VPC, IAM).
- Build and maintain agentic harnesses: tool definitions, MCP servers, context management, memory, retrieval, sub-agent orchestration and human-in-the-loop checkpoints.
- Own the production characteristics that matter — latency, token cost, failure modes, retries, rate limits and graceful degradation.
- Build the evaluation and observability layer: tracing, regression suites, and evals that tell us whether a change actually improved anything.
- Implement guardrails, identity and access controls, and data handling that stand up to enterprise security review and GDPR scrutiny.
- Ship infrastructure as code, CI/CD and repeatable deployment patterns we can reuse across clients.
- Work directly with client engineering teams and leave them able to run what we built.
What you'll need
Essential
- Expert Python. You write production code — typed, tested, reviewed — not notebooks.
- Strong cloud infrastructure background on AWS, with real depth in IAM, networking and containerised workloads. Infrastructure as code (Terraform or CDK) as a default habit.
- Hands-on production experience with Amazon Bedrock and Bedrock AgentCore.
- Demonstrable experience building agentic harnesses — with Claude, Kiro, or comparable frameworks and agentic coding tools. We care about what you have shipped and what broke, not which vendor's logo is on it.
- A working understanding of where LLM systems fail: context degradation, tool-call errors, prompt injection, non-determinism, and runaway cost.
- Clear written and verbal communication. You will be in front of clients.
Desirable
- Model Context Protocol (MCP) server development.
- Eval frameworks and LLM observability tooling.
- Data engineering background — pipelines, warehouses, lakehouse patterns.
- Prior consultancy or client-facing delivery experience.
- Exposure to regulated environments (financial services, health, public sector).
What we offer
- Competitive salary
- Hybrid working from our Dublin base.
- Genuine variety of work and direct access to senior clients from day one.
- A team small enough that your decisions carry.