DevSecOps AI Engineer

1. Agentic Automation (The Builder Role)

• Build Operations Agents: Develop intelligent agents using Vertex AI Agent Builder, LangChain, and Python.

• Infrastructure Interaction: Design Function Calling capabilities that allow Gemini models to securely interact with our infrastructure (e.g., Agent, check why this pod crashed and fetch the logs).

• RAG Implementation: Build Retrieval-Augmented Generation pipelines to ground agents in our internal runbooks and architecture documentation.

2. AI & MLOps Pipeline Security

• Secure the Supply Chain: Architect hardened MLOps pipelines using Vertex AI and Kubeflow, ensuring strict chain-of-custody for training data and model artifacts.

• LLM Guardrails: Implement security controls for Generative AI endpoints to prevent Prompt Injection, Jailbreaking, and PII leakage (using tools like NVIDIA NeMo or custom GCP logic).

3. GCP Infrastructure & Governance

• Infrastructure as Code: Manage ephemeral training environments and persistent inference clusters (GKE Autopilot) using Terraform.

• Policy & Isolation: Implement VPC Service Controls and Organization Policies to create security perimeters around sensitive BigQuery datasets.

• ML-Specific CI/CD: Build pipelines (Cloud Build/GitHub Actions) that strictly automate model evaluation and bias detection before deployment.

4. Security Operations (SecOps)

• Vulnerability Management: Integrate container scanning (Artifact Registry) and SAST/DAST into the ML workflow.

• Identity Architecture: Design Least Privilege access models for both humans and AI agents using Workload Identity Federation.

Technical Requirements:

• Cloud Platform: 4+ years of hands-on experience with Google Cloud Platform (GCP), specifically Vertex AI, GKE, BigQuery, and IAM.

• AI Development: Strong proficiency in Python with experience building agents/apps using LangChain or Vertex AI APIs.

• DevOps Tooling: Expert-level Terraform skills and proficiency with GitHub Actions.

• Containerization: Deep understanding of Docker and Kubernetes (including GPU resource management).