Carl van Laresenior

Led infrastructure and release engineering for a multi-tenant enterprise AI SaaS platform on Azure, scaling it from proof of concept to production for enterprise clients. Architected and ran an agentic AI platform with 10+ agents built on MCP, A2A and AG-GUI, routing LLM traffic through a LiteLLM gateway across model deployments and managing per-provider quotas, handling ~2M requests. Built evaluation pipelines per agent and led creation of a CI-gated evaluation suite using golden-set tests targeting a 95% release pass rate. Owned observability and incident response against a 99.9% availability target, with zonal and regional failover and runbooks aligned to RTO/RPO goals. Implemented usage-based chargeback for LLM costs as part of a FinOps model by token tracking and per-client rate limits. Delivered tenant isolation via infrastructure, configuration, and data, supporting routine onboarding changes.

Modernized platform delivery within a client engineering team by rebuilding CI/CD in GitHub Actions, cutting delivery time by 35% using Checkov, Snyk, and Trivy scans and automating PostgreSQL migrations in the pipeline. Reduced non-production cloud spend by 30% using scheduled shutdowns, lifecycle policies, and removal of orphaned resources. Replaced hard-coded credentials with federated OIDC credentials and managed identities across pipelines and workloads, tightening edge security. Supported infrastructure build-out in Terraform with consistent RBAC, tagging, and policies across environments. Improved traceability and incident investigation using Azure Monitor and Log Analytics.

Moved application delivery to GitOps using Argo CD and Helm for consistent, traceable AKS releases. Reduced pipeline runtime by 30% and infrastructure costs by 25% through caching, dynamic triggers, and out-of-hours shutdowns for non-production. Delivered an Azure tenant migration with zero downtime. Wrote runbooks to cut onboarding time for new engineers by 40%. Mentored early-career engineers through coding bootcamp partnerships. Brought managed Azure resources under Terraform and Terragrunt so infrastructure and access controls were managed in one place. Added high-availability improvements, security scanning (Snyk, SonarQube) and Azure Monitor alerting to surface issues earlier.