Role

Senior AI Engineer

Location: United States (Remote)

Reports to: AI Lead

Employment Type: Full-time, Exempt

Direct Reports: None

What is the role?

Senior AI Engineer builds and ships AI enabled software. This is a hands-on build role on Howden's US AI Engineering team: writing the agents, the pipelines, the integrations, and the infrastructure as code that puts AI capability into production on Azure. This is a software engineering role first. The work is production code in Python, Node.js and TypeScript, and .NET (C#), sitting on top of Azure services and Howden data.

You will work from solution designs and integration patterns set by the AI Lead and enterprise architecture, and you will own the implementation end to end. Expect to spend most of your time in code, in Terraform, and in CI/CD pipelines, with a heavy emphasis on agent-assisted development: using coding agents and AI development tooling to move faster than a conventional engineering pace and knowing when to trust the output and when to rewrite it.

This role is self-directed. You will be handed an outcome and a rough shape, and you are expected to break it down, sequence it, unblock yourself, and deliver production-quality work collaborating with other team members.

What success looks like

• AI agents and services running in production on Azure, built by you, with tests, telemetry, and deployment automation attached.

• Terraform and pipeline code that provisions and deploys AI workloads repeatably across environments with no manual steps.

• Vendor AI platforms wired into Howden systems through working integrations, not slide decks.

• Knowledge bases and retrieval layers that agents can rely on, with retrieval quality measured against real queries rather than assumed.

• Agent-assisted development practices are used daily and shared with the team as working examples: prompts, harnesses, evaluation scripts, and repository conventions that make coding agents productive on our codebase.

• Work delivered from a stated outcome with minimal direction, escalating early when something genuinely needs a decision above your level.

• Design input that changes outcomes: options, trade-offs, and working spikes brought to the AI Lead and architecture before decisions are locked.

What will you be doing?

AI Agent & Application Development

• Build production AI agents and conversational solutions on Azure using Azure OpenAI Service, Azure AI Foundry, Azure Bot Framework, and agent orchestration frameworks.

• Implement multi-agent orchestration, tool and function calling, retrieval-augmented generation, and context and memory management in code.

• Build retrieval pipelines and knowledge bases using Azure AI Search and vector stores, including chunking strategies, embedding pipelines, and document processing.

• Write evaluation harnesses and regression tests for AI behavior: golden datasets, scoring scripts, and automated checks that run in CI before anything ships.

• Develop RESTful and event-driven APIs and services that expose AI capability to internal applications, using Azure Functions, Container Apps, API Management, and Service Bus.

• Refactor and harden prototypes into supportable production code with error handling, retries, rate limiting, and cost controls.

Data & Knowledge Engineering

• Build and maintain the knowledge layer behind our AI systems: source profiling, extraction, cleansing, normalization, and enrichment across structured and unstructured data.

• Model the domain. Define taxonomies, ontologies, entity and relationship models, and metadata schemas that give agents a consistent view of insurance data such as clients, policies, carriers, submissions, and claims.

• Design and tune retrieval quality end to end chunking strategy, embedding choice, hybrid and semantic search, metadata filtering, reranking, and groundedness evaluation.

• Build ingestion and refresh pipelines that keep knowledge bases current, with lineage, versioning, change detection, and reconciliation against source systems.

• Analyze system and usage data to find where AI is failing query logs, retrieval hit rates, groundedness scores, cost per interaction, and user feedback. Act on what the data shows.

• Write the SQL, transformations, and analysis needed to answer your own data questions rather than waiting on another team.

Agent-Assisted Development

• Use coding agents and AI development tooling (Claude Code, GitHub Copilot, and similar) as a primary part of your daily workflow for implementation, refactoring, test generation, and debugging.

• Build and maintain the scaffolding that makes agent-assisted development work on our repositories: context files, tool definitions, repository conventions, task decomposition patterns, and reusable prompt assets.

• Apply engineering judgment to agent output. Review generated code as rigorously as human-written code and know where the tooling saves hours and where it creates clean up work.

• Automate repetitive engineering work with scripted agents: migrations, test backfill, documentation generation, dependency upgrades, and integration scaffolding.

• Share working patterns with the team through examples in the codebase rather than through process documents.

Infrastructure as Code & Deployment Automation

• Write and maintain Terraform for Azure AI workloads: Azure OpenAI deployments, AI Search, Cosmos DB, Storage, networking, Key Vault, managed identities, and role assignments.

• Own module structure, state management, workspace and environment separation, variable and secret handling, and drift detection for the infrastructure you build.

• Build CI/CD pipelines in Azure DevOps or GitHub Actions covering plan and apply gates, automated testing, container builds, environment promotion, and rollback.

• Containerize AI services and deploy Azure Container Apps, Container Instances, or AKS with sensible scaling and resource configuration.

• Keep environments reproducible. Anything created by hand in the portal gets replaced by code.

Vendor Platform & API Integration

• Implement integrations between vendor AI platforms and Howden systems using vendor APIs, SDKs, webhooks, and connectors.

• Handle authentication and authorization with Entra ID, OAuth 2.0, managed identities, and API keys, with secrets managed through Key Vault.

• Build and test data flows between third-party platforms and internal systems, including error handling, replay, and reconciliation.

• Support technical evaluation and proof-of-concept work on vendor platforms by building working spikes against real APIs.

• Ship AI capability into Microsoft Teams, web channels, and other surfaces where our users already work.

Observability & Production Support

• Instrument AI services with Azure Monitor, Application Insights, and Log Analytics, including distributed tracing, token and cost telemetry, and structured logging.

• Build dashboards and alerts for latency, error rates, throughput, model usage, and cost.

• Participate in production support for AI systems: triage, root cause analysis, fixes, and post-incident follow-through.

• Write runbooks and troubleshooting notes for the systems you build so others can operate them.

Responsible AI in Practice

• Implement content filtering, safety guardrails, and Azure AI Content Safety configurations in the solutions you build.

• Implement PII detection and redaction in AI interactions and data pipelines.

• Build enterprise security and data privacy standards: access controls, encryption, network isolation, and audit logging.

Solution Design Contribution

• Contribute to solution designs alongside the AI Lead and enterprise architecture: bring options, trade-offs, cost and latency implications, and a recommendation grounded in what you have built.

• Build spikes and reference implementations that prove or kill a design approach before the team commits to it.

• Break approved designs into technical work: component boundaries, data contracts, interfaces, sequencing, and realistic estimates.

• Review vendor and internal integration designs and flag gaps in security, scale, cost, or supportability early.

• Document the as-built design: what shipped, where it deviated from the original approach, and why.

Collaboration

• Work inside an Agile team with product managers, engineers, architecture, and security to plan, refine, and deliver iteratively.