• Built a one-click Azure infrastructure bootstrap for a major enterprise client, enhancing DRP readiness to meet 99.9% uptime requirements

• Managed FinOps operations using DoIT and native GCP, Azure, and AWS cost tools to monitor spend, analyze trends, and detect anomalies

• Led Helm chart standardization and migration efforts via Jenkins migration pipelines, tracking progress with Honeycomb dashboards

• Cut MongoDB production costs by 56% ($22.8k/year) by optimizing database storage and reclaiming unused space with zero downtime

• Saved 7 weeks/year by automating company-wide Helm changes via centralized base charts, cutting deployment time from months to <1 hour

• Ensured GDPR compliance with Terraform-managed GCS archives using TTL rules and audit-proof access controls for deleted medical data

• Developed internal tools to enhance disaster recovery, including exporting PostgreSQL backups to external GCS buckets

• Handled on-call incidents for cloud and Kubernetes with PagerDuty, kubectl, Honeycomb tracing, and E2E testing

• Enabled regulatory compliance in Canada by mentoring the local infra team on secure CI/CD pipelines and access governance

• Improved infrastructure robustness with Terraform, including prevention of accidental destruction and smooth replacement of resources

• Cut Terraform provisioning time by 89% (18 → 2 mins) for 134 customer-specific IaC folders through selective provisioning, saving $1.8k/year

• Reduced MongoDB IAM-related support tickets by 60% by integrating IAM with Terraform, enabling self-service access via approval workflows

• Consolidated multi-provider (GCP, AWS, & Azure) Terraform code for a global health company

• Developed an ad-hoc license detector in Java for the EU-funded project Fasten, integrated into their pipeline (Kafka on Kubernetes on-premise)

• Cut Kubernetes costs 35% by profiling deployments with Goldilocks, right-sizing resources, and reducing nodes by 40%

• Cut Docker build times 79% (from 104 to 22 seconds) by optimizing Maven multi-stage builds, caching artifacts, and enabling BuildKit

• Enabled a team of 3 developers to work concurrently and accelerate delivery by parallelizing the pipeline with selective task execution

• Built a flow-based cluster scheduler aware of network resources; helped extend its Scala-based simulator and refine the solver's cost model

• The simulator featured in the 2021 ASPLOS paper "Switches for HIRE: Resource Scheduling for Data Center In-Network Computing"