Senior AI Platform Engineer

AddanEx JOB POSTING:
Title: Senior AI Platform Engineer
Location: Remote
Onsite / Remote: 100% Remote within EU
Type: Contract
Start Date: ASAP / September – October
Duration: 6 Months + Extensions
Utilisation: 40 Hours per week
Languages: English

Job Summary / Key Skills:

• Has set up an AI governance platform end to end before: gateway, agent repository or registry, policy enforcement, usage and cost telemetry, access provisioning, whatever their build covered. This engagement is a second build, not a first. The candidate should be able to walk through the platform they built, what it governed, who used it, and what they would do differently
• Production Kubernetes: has run containerised workloads other teams depended on
• API gateway or LLM proxy engineering: APIM, Kong, LiteLLM or equivalent
• Practical LLM application work: API integration, token cost behaviour, evaluation, RAG patterns
• Production observability ownership: OpenTelemetry, Grafana, Log Analytics or equivalent
• Strong backend engineering experience preferably with GO; IaC and CI/CD (Terraform, GitHub Actions or Azure Pipelines) (Python and/or TypeScript can be considered as an alternative)
• Identity and access engineering: SSO, RBAC, service principals
• Has built an internal platform or developer service that other teams adopted voluntarily
• Deliberately not required: ML/data-science depth. This is a platform role, not a modelling role.

Task:
Our client is setting up a central AI governance platform, gateway, agent repository, cost telemetry, golden paths, provisioning automation, serving developers across a decentralised group. The platform's promise to its users is cheaper, faster, and more reliable than calling AI vendors directly; governance and cost attribution ride along by default rather than being the selling point.

The Job:

• Build and operate the AI gateway: SSO, per-request logging, data-classification tagging, policy enforcement, model routing, on Azure (Azure OpenAI / AI Foundry, Entra ID, APIM)
• Stand up the agent repository: a governed catalogue of approved agents with owner, permissions and data scope on each, the control point for everything the gateway serves
• Build the cost telemetry pipeline: every request attributable to a cost centre and a use case Automate access provisioning and seat lifecycle for each AI tool licence the group adopts, current and future
• Produce developer golden paths: RAG template, agent template, evaluation harness
• Establish observability and runbooks so L1 support can operate the platform's routine work

Delivered by month six:

1. Gateway v1 in production for one pilot division, with logging and cost attribution on every request
2. Agent repository v1 live: every agent served by the gateway is registered, owned and permissioned
3. Live cost telemetry: request cost centre + use case, feeding a monthly report
4. Seat reclamation automated for the largest licence pool
5. Three golden-path templates published and used by at least one team outside the pilot
6. Runbooks covering the top five ticket types, executable by L1 without escalation