Senior MLOps Engineer – Agentic Versioning

We are looking for a Senior MLOps Engineer to design and implement enterprise-grade versioning and lifecycle management for AI agents. The role focuses on treating an agent, including its container image, prompt configuration, tool bindings, and policy associations, as a single immutable versioned unit, with controlled promotion, validation, and rollback across environments.

Our client is a large global enterprise operating across multiple markets, with a complex technology landscape and a strong focus on digital transformation and innovation. The organization is actively investing in modern cloud, data, and AI capabilities to enable scalable, secure, and highly automated solutions across its business.

The project is focused on building an enterprise-grade AI Agent Platform from the ground up. The platform will provide a standardized foundation for developing, deploying, orchestrating, securing, and observing AI agents across multiple teams and use cases.

The initiative covers the full agent lifecycle and brings together agent orchestration, observability, security, governance, integrations, evaluation, and platform engineering. Engineers joining the project will have an opportunity to influence key architectural and technical decisions and contribute to a new platform rather than maintaining an existing solution.

• Design and implement an agentic versioning model covering container images, prompt configurations, tool bindings, and policy associations as immutable versioned artifacts.
• Implement version management for AWS AgentCore Runtime, including automatic version creation and endpoint management.
• Build agent promotion workflows across dev, QA, and production, with evaluation and quality gates before production deployment.
• Define and implement rollback mechanisms for restoring previous agent versions and validating their continued functionality.
• Manage agent container images in AWS ECR, including ARM64 tagging, lifecycle policies, image signing, and immutability controls.
• Integrate versioning workflows with an Agent Registry, automatically registering new agent versions and associated metadata.
• Capture and maintain metadata such as evaluation results, promotion status, and version information.
• Design CI/CD pipelines that enforce quality and evaluation gates for AI systems.
• Support version comparison and regression detection across agent releases.

• Agentic versioning model — agent container image + prompt configuration + tool bindings + policy associations as immutable unit

• AWS AgentCore Runtime versioning — automatic version creation on image push, endpoint management (DEFAULT, named)

• Agent promotion workflow — dev → qa → prod with evaluation gates enforcing quality before production

• Agent rollback patterns — restoring previous agent versions with validation that old version still works

• AWS ECR for agent images — ARM64 tagging, lifecycle policies, image signing for agent immutability

• Agent Registry integration — auto-registering agent versions with metadata (evaluation results, promotion status)

• 5+ years MLOps or DevOps engineering with AI/ML systems

• Versioning and lifecycle management for LLM agents or ML models

• Immutable artifact versioning at enterprise scale (ECR, model registries)

• CI/CD pipeline design with quality gates for AI systems

• AWS AgentCore Runtime versioning hands-on experience

• Agent evaluation frameworks (Langfuse, LangSmith, PromptLayer)

• Multi-environment agent promotion patterns

• Agent version comparison and regression detection