1. Infrastructure as Code (IaC) & Cloud Architecture

• Design, build, and maintain production GCP environments using modular, reusable Terraform (and Terragrunt) code.

• Manage multi-environment cloud resources across VPC networking, Private Service Connect, Cloud Interconnect, IAM, KMS, and Cloud Storage.

• Enforce security best practices, least-privilege IAM policies, Workload Identity Federation, and automated compliance auditing.

2. LLMOps & GenAI Platform Engineering

• Build automated deployment, scaling, and operational infrastructure for generative AI models hosted on Vertex AI, GKE (Google Kubernetes Engine), or hybrid inference servers.

• Implement or Understanding end-to-end LLM lifecycle pipelines: prompt management, fine-tuning workloads, automated retrieval-augmented generation (RAG) components, and evaluation harness pipelines.

• Deploy, manage, and scale vector databases (e.g., Vertex AI Vector Search, Pinecone, Qdrant, pgvector) with proper indexing, backup, and latency monitoring.

• Architect and optimize model-serving engines.

3. CI/CD & Pipeline Automation

• Build and maintain scalable CI/CD pipelines using Cloud Build, GitHub Actions, or GitLab CI for both microservices and machine learning models.

• Establish automated testing for infrastructure changes (e.g., tflint, tfsec, terratest) and model release pipelines (canary deployments, blue/green rollouts, shadow traffic).

• Package and deploy containerized AI/ML applications using Docker, Artifact Registry, and Helm charts on GKE.

4. Observability, Guardrails & Reliability

• Set up deep observability stacks using Google Cloud Monitoring, OpenTelemetry, Prometheus, and Grafana to track standard SRE metrics (SLIs/SLOs, latency, throughput).

• Implement LLM-specific telemetry: token consumption rates, inference latency, model drift, cost-per-query, and hallucinations/groundedness metrics using tools such as LangSmith, Phoenix/Arize, or Vertex AI Model Monitoring.

• Integrate safety guardrails, prompt sanitization, rate limiting, and fallback/resiliency mechanisms across inference endpoints.

5. FinOps & Compute Optimization

• Monitor and optimize cloud spending for compute-intensive workloads, specifically LLM quota monitoring, spot instances, and dynamic autoscaling.

• Partner with engineering and product teams to project infrastructure costs and establish budget s and quota guardrails.