Principal AI Engineer DevOps

As an AI Systems Engineer, you will own the infrastructure, deployment, and reliability of production-grade agentic AI systems that power the next

generation of intelligent automation at project. Reporting to the Global Head of AI Strategy, Policy, and Governance, you will operate at the intersection of cloud infrastructure engineering and AI systems, provisioning, securing, and scaling the AWS environments that make our AI pipelines run reliably and cost-efficiently.

What makes this role interesting?

Ownership: Own real infrastructure. This is not a role where you configure demos, you will provision, harden, and operate the AWS environments that run client AI agents at scale.

Frontier Stack: Work directly with AWS Bedrock AgentCore, a capability that most cloud engineers are only beginning to encounter. You’ll be among the first teams operationalizing it in production.

High Impact: Small team, direct line to executive leadership, real decisions.

Here’s how you’ll be making an impact:

IaC and Cloud Provisioning: Design and maintain infrastructure-as-code (Terraform) for all AI workloads: Bedrock agents, AgentCore runtimes, Lambda functions, API Gateway endpoints, and supporting data services.

AWS Bedrock AgentCore: Own deployment, configuration, and scaling of AWS Bedrock AgentCore environments, including model invocation routing, session management, and memory backends.

Security and IAM: Architect secure, least-privilege IAM policies for agent runtimes, MCP integrations, cross-account access patterns, and service-to-service authentication.

CI/CD Pipelines: Build and maintain CI/CD pipelines (GitHub Actions, AWS CodePipeline) for agentic workloads, covering infrastructure changes, Lambda deployments, and agent configuration updates.

Observability and Cost Management: Instrument AI systems with OpenTelemetry, CloudWatch, and X-Ray. Own observability strategy: traces, metrics, cost dashboards, and alerting across agent pipelines.

Networking: Manage networking topology for AI workloads: VPC design, PrivateLink, security groups, and egress controls to ensure data never leaves governed boundaries unexpectedly.

Data Infrastructure: Build and operate the data infrastructure supporting agents: OpenSearch for semantic search and vector retrieval, and S3 lifecycle policies for knowledge artifact storage.

Collaboration: Collaborate with AI engineers on MCP server deployments, containerized agent runtimes (ECS/Fargate), and performance tuning to hit latency and cost targets.

Job Responsibilities:

The ideal candidate will have hands-on experience owning production infrastructure, with deep proficiency across AWS and the modern cloud engineering stack:

Infrastructure as Code: You have strong hands-on Terraform experience. You version-control infrastructure the same way you version-control code.

AWS Platform Depth: Deep AWS experience across Bedrock (model invocation, agents), Lambda, API Gateway, ECS/Fargate, App Runner, IAM, VPC, S3, CloudWatch, and X-Ray. Bedrock AgentCore exposure is a strong plus.

Security Mindset: You understand the AWS IAM model deeply: roles, policies, SCPs, permission boundaries, and cross-account trust. You instinctively scope to least privilege.

CI/CD: You build pipelines that deploy infrastructure and application code reliably. GitHub Actions, CodePipeline, or equivalent. You know how to roll back safely.

Observability: Experience with OpenTelemetry, CloudWatch Logs Insights, and distributed tracing. You care about visibility into what AI systems are actually doing and what they cost.

Python: Comfortable with Python for scripting, Lambda functions, and lightweight automation. You don’t need to be an ML engineer, but you can read and modify agent code.

Containers: Comfortable with containerized workloads (Docker, ECS, Fargate, App Runner). Experience running long-running agent processes or streaming inference endpoints is a plus.

Data Infrastructure: You understand vector search infrastructure and are comfortable operating OpenSearch clusters for semantic retrieval. Experience with RAG pipeline data stores preferred.