AI DEVOPS ENGINEER | GITLAB DUO | ATLASSIAN ROVO | CI/CD & AGENTIC WORKFLOWS

• Designed an end-to-end AI-assisted delivery workflow connecting Jira and Confluence context with GitLab, defining a clean

handover from approved user story to technical specification and development kick-off.

• Built intake-agent patterns that accept Jira stories, validate required fields, pull repository and documentation context, and

assemble bounded context packages for downstream LLM processing.

• Designed specification-agent logic to derive affected components, interfaces, data-model changes, implementation steps, risk

areas, and test cases, with explicit approval before code generation begins.

• Implemented GitLab automation for branch creation, code-scaffold generation, initial unit/integration tests, pipeline execution,

merge-request creation, and automated merge-request summaries.

• Integrated GitLab APIs and webhooks for repository events, pipeline status, merge-request lifecycle, approvals, and

synchronization of development status back to Jira.

• Engineered GitLab CI/CD pipelines and runner execution patterns with reusable templates, job dependencies, artifacts, caching,

environment controls, failure handling, and scalable pipeline composition.

• Applied GitLab permission-model controls including service accounts, scoped access tokens, protected branches, mandatory

review, least-privilege access, secret handling, and auditable automation events.

• Worked with GitLab Duo capabilities such as Chat, Code Suggestions, and agentic workflow concepts to accelerate specification,

coding, test generation, and developer review without bypassing engineering controls.

• Defined operational guardrails for AI-generated code: mandatory human review, restricted write permissions, approval gates,

pipeline quality checks, security scanning, traceability, and controlled rollback paths.

• Produced demo scenarios, technical documentation, operating guidance, and knowledge-transfer material so the prototype could

progress toward repeatable enterprise-scale adoption rather than remain a one-off proof of concept.