AI Platform Architect

At Signature Aviation we are modernizing operations and customer experience through advanced data platforms and artificial intelligence. We are hiring an AI Platform Architect to design, build, and operate the Azure foundation that AI initiatives across the enterprise run on: the AI landing zone, the Azure API Management gateway that fronts all model access, the model evaluation framework, AI Gateway and the governed Azure sandboxes where teams experiment. You will define the standards rather than inherit them, and you will prove them by shipping real use cases on the platform you build.

Working model — forward deployed: you embed with the business unit or engineering team that owns a use case, build inside the environment and manage the release process, hand back a running system, and generalize what you built into platform capability between deployments.

Essential Duties and Responsibilities
• Design and build the shared enterprise AI platform and reference architectures for LLM applications, agentic systems and multi-agent solutions using LangGraph, Microsoft Agent Framework, Semantic Kernel or equivalent technologies.
• Establish reusable patterns for agent state, checkpointing, persistence, human-in-the-loop, long-running execution, memory, context management and single- or multi-agent orchestration.
• Define standards for MCP, function calling and enterprise tool integration, including authentication, authorization, schema validation, policy enforcement and auditability.
• Build and operate the enterprise AI gateway/model proxy using Azure API Management or equivalent technology, supporting provider-agnostic model access, routing, fallback, retries, quotas, caching, token metering and cost attribution.
• Build shared retrieval capabilities and patterns for ingestion, embeddings, vector and hybrid search, metadata and reranking using Azure AI Search, Databricks Vector Search, pgvector, Cosmos DB or equivalent technologies.
• Build common model and agent evaluation and observability capabilities using golden datasets, LLM-as-judge, regression testing, agent trajectory evaluation and end-to-end tracing across applications, models, retrieval and tools.
• Build and operate the Azure AI foundation using secure networking, Entra ID, RBAC, Key Vault, Azure Policy, AKS, Azure Container Apps, Azure Functions, infrastructure as code and CI/CD.
• Build governed AI sandboxes and platform controls for data access, PII, prompt injection, content safety, auditability, cost management and appropriate constraints on agent autonomy.
• Work as a Forward Deployed Engineer with business and engineering teams on real AI use cases, helping move them from experimentation into production and using those deployments to identify gaps in the shared platform.
• Generalize lessons from deployments into reusable platform services, SDKs, templates and engineering standards so teams can build and operate AI capabilities without rebuilding common infrastructure.

Minimum Education and/or Experience

• 10+ years in software, cloud, or platform engineering, spent primarily hands-on.

• You have built an AI or ML platform that other teams used in production — you owned it, or a major component of it, including workloads running in live operational settings.

• Kubernetes and distributed systems depth, with strong cloud networking, security architecture, and reliability engineering practice.

• Azure is required: Microsoft Foundry (formerly Azure AI Foundry) including Azure OpenAI models — projects and deployments, the Foundry Models catalog, the Foundry Agent Service, the Microsoft Agent Framework, and Azure AI Content Safety.

• AKS or Azure Container Apps, Bicep or Terraform, and Azure DevOps or GitHub Actions — pipelines and modules you wrote, not ones you inherited.

• You have built or operated an LLM/AI gateway: token metering, model routing, quota enforcement, caching. Azure API Management preferred; LiteLLM, Kong, Apigee, or a bespoke proxy considered on merit.

• Vector and hybrid retrieval — Azure AI Search, Cosmos DB, or pgvector — plus embedding and ingestion pipelines that stayed correct as source data changed.

• You have built model evaluation from scratch — datasets, metrics, judges, regression suites — and can explain how you knew a model change was an improvement.

• Microsoft Entra ID, Azure RBAC, Key Vault, and Private Link in a production estate.

• Strong Python, plus enough TypeScript to build and debug an application front to back.

• You have delivered while embedded with a business or engineering team, accountable for the outcome rather than for a component handed across a boundary.

• Enterprise judgment on data classification, PII handling, access control, auditability, and responsible AI.

• Bachelor's degree in Computer Science, Engineering, or a related technical field, or equivalent practical experience.

Preferred Qualifications

• Strong LangGraph, multi-agent or agent-runtime experience.

• MCP, enterprise tool gateways and identity-aware tool execution.

• Agent memory and context management architectures.

• Azure API Management, LiteLLM, Kong, Portkey or similar AI gateway/model proxy technologies.

• Microsoft Foundry, Azure OpenAI, Azure AI Search, Azure Databricks or Microsoft Fabric.

• OpenTelemetry, Langfuse, LangSmith, Ragas, Promptfoo or Azure AI Evaluation SDK.

• Temporal, Azure Durable Functions or similar durable workflow technologies.

• MLflow and model lifecycle management.

• Commercial and open-source model deployment, including GPU workloads.

• Experience in a Forward Deployed Engineer, platform consulting, solutions engineering, startup or internal innovation environment.

• Aviation, logistics, mobility or other large-scale operational environments.

Additional Information

This is an applied engineering role: you will be in the code and the pipelines every week. It is not a research role, and it does not cover building statistical or forecasting models — you build and operate the platforms and serving paths their outputs run on.