Vultr
Role Held: Sr. Director of Platform Engineering
Duration: 1 year
Brief Description: Vultr is a public cloud provider that over its history has focused on several niches within the hosting and colocation space. My interest in Vultr was to join and help mature their operations and grow my knowledge around AI infrastructure including high speed networking, GPU workloads for inferencing and training AI workloads.
Key Achievements:
- Led and developed the initial formation of the platform engineering team at Vultr. Provided patterns and guidance for CI/CD, Preproduction environments, and worked to close business impacting gaps for K8s, Virtualization and GPU Workloads (Inference and Training) through Observability implementation and data analysis around runtime and provisioning user experiences.
- Was part of a cross functional team involved with GTM Strategy for GPU Acceleration. My team along with the dedicated AI/ML infrastructure team performed workload optimization for single GPU and Clusters for inference and training workloads that were in alignment with vendor reference architectures. We optimized fractional, whole card, and networked GPU clusters (Nvidia vGPU, RoCE/Infiniband) telemetry and operations.
- Security & Compliance Leadership: Directed engineering and audit efforts to secure ISO 27001 and ISO 20000 certifications, establishing best-in-class security and service management standards. Additionally, assisted in overhauling PCI compliance and alignment with the architecture deployed to allow for successful audits.
- Provisioning Automation & SLA Optimization: Designed a provisioning observability pipeline and jobs to capture ongoing failure rate, time to provision and other key metrics to assist with systematically targeting improves that; reduced error rate by ~15% in aggregate, lowering time-to-provision by 40%, and improving the trustworthiness of our platform.
More project breakdowns may be added in the future. However, out of respect for Vultr, my most recent employer, I’m keeping this section high level.