Remote job
AI Engineer - Agent Infrastructure & AI Workflow
Job details
About this role
Role overview Build the shared cloud infrastructure that runs long-lived AI agent workflows for an AI-assisted hardware modeling pipeline and internal software development tools. You will own orchestration, model access, evaluation, observability, and secure deployment so teams can operate these workflows as dependable services.
Responsibilities - Design AWS orchestration for long-running agent jobs, including checkpoints, retries, artifacts, manifests, and version tracking. - Route inference across managed and self-hosted models; add model judging, semantic caching, and controls for latency and token costs. - Provide evaluation and tracing tools so teams can compare pipeline stages, prompts, and model versions. - Build and maintain an internal AI gateway, agent runtime, and integrations for code review, ticket-to-merge workflows, and knowledge retrieval. - Maintain CI/CD and infrastructure as code, applying secure secrets handling and least-privilege access. - Track reliability, cost, latency, and failure patterns with dashboards, alerts, and runbooks; coordinate interfaces with teams responsible for external APIs and MCP tools.
Requirements - Practical experience with AWS services such as Bedrock, Lambda, ECS or EKS, S3, and Step Functions. - Experience with generative AI infrastructure, including LLM integrations, agent frameworks, MCP, semantic caching, and evaluation. - Familiarity with Kubernetes deployments, GitHub or GitLab workflows, secrets management, and secure delivery. - Knowledge of observability tools and log aggregation, with the ability to write maintainable code and documentation. - Strong communication and collaboration skills, including explaining system behavior, limitations, risks, and trade-offs to stakeholders. - Comfort with changing requirements, broad ownership, and occasional travel for client workshops where needed.
Nice to have Experience improving agent behavior through runtime design, tools, workflows, context, and safeguards; familiarity with production AI systems and client-facing delivery is useful.