Max Z.

About Max Z.

Senior SDE with 5 years of experience. Developed and managed AI infrastructure at OCI including NVIDIA GB200 GPU clusters to optimize utilization for large scale AI models training, Kubernetes infrastructure orchestration, workflow automation, cloud-native DevOps and CI/CD, focused on large scale GPU-based infrastructure.

Experience

1. Senior Software Engineer

   Oracle

   Mar 2024 — Present · Seattle, WA, US

   Developed infrastructure automation workflow for clients such as NVIDIA to self-service configuration and provisioning of GPU-compute resources for AI training and inference use cases, and maximize cluster operation efficiency and utilization.2. Designed a fleet orchestration workflow, with configurable rules logic and integration with ML models to decide and trigger automated fleet provisioning and management actions in response to health, efficiency and user-initiated instructions.3. Developed metrics data pipeline based on cluster, container and service metrics. Defined a single canonical data model, and executed near-real time aggregation and processing data across more than total instances.4. Developed instance monitoring and action daemons to support barebones hardware control across OCI’s instance types.5. Developed a GenAI powered fleet optimization opportunity advisor agent to digest large amounts of configuration, health and utilization data and generate natural language summary about possible inefficiency and oppurtunities.Implemented context engineering based on past user actions and metric snapshot, SOP and best practice documents, to improve suggestion relevance. Proposed a multi-agent discussion mechanism to reduce hallucination.6. Implemented a full-stack fleet management console with visualization, interactive controls and embedded agentic UX.7. Proposed a mechanism to dynamically prioritize request and action based on type, scope, and resource dependency.8. Improved concurrency, utilization and improved critical task execution SLA, improved overall utilization from 86% to 92%9. Proposed and implemented a data reconciliation scanner to fix fleet metadata inconsistency issues. Mitigated operational pain due to race conditions and failures during fleet provisioning. Reduced weekly operational effort by 40+ hours.10. Setup CI/CD production environment resource automation pipelines with Kubernetes Terraform and Herds.

Education

• Northeastern University

  Master of Computer Science, Computer Science