GPU Cluster Orchestration for LLM Training
Summary
Configured and automated large-scale Slurm-based GPU clusters to support fine-tuning and inference of open-source LLMs.
About
Highly accomplished DevOps & AI Infrastructure Engineer with over 5 years of experience specializing in scaling GPU-based LLM training clusters (100+ GPUs) and automating multi-cloud MLOps pipelines across GCP and Azure. Proven leader in reducing deployment times by 60% and optimizing GPU utilization through expert application of Terraform, Helm, and Kubeflow. Passionate about architecting robust, scalable AI infrastructure that seamlessly bridges model research and production reliability.
Work
Jukshio
|DevOps Engineer
Hyderabad, Telangana, India
→
Summary
Led the design, implementation, and optimization of robust DevOps and AI infrastructure, ensuring high performance and scalability for large-scale LLM operations.
Highlights
Orchestrated and optimized Slurm-based GPU clusters for large-scale LLM fine-tuning, supporting over 100+ GPUs to enhance model development capabilities.
Deployed scalable Kubernetes inference services utilizing SGLang and vLLM, significantly improving latency and throughput for critical customer workloads.
Automated infrastructure provisioning across multiple cloud environments using Terraform, Helm, and Kustomize, cutting setup time by 40%.
Designed and implemented multi-stage CI/CD pipelines in GitLab CI, reducing release cycles by 60% through integrated build, test, security scan, and deployment automation.
Enhanced compliance and security posture by implementing DevSecOps practices with SonarQube, Trivy, and OWASP WAF across all development stages.
Developed and automated a Kubeflow-based MLOps pipeline, effectively supporting 20+ model training jobs across diverse GPU clusters.
Managed and optimized Prometheus and Grafana observability stack, reducing time-to-detect incidents by 40% and enhancing system reliability.
Successfully migrated critical workloads from Azure to GCP, improving system reliability and reducing overall cloud costs by 25%.
Wipro
|Test Engineer
Chennai, Tamil Nadu, India
→
Summary
Contributed to quality assurance processes by automating regression testing and developing data validation utilities to accelerate QA cycles.
Highlights
Automated comprehensive regression testing using Selenium WebDriver (Java), significantly improving testing efficiency and accuracy.
Developed robust data validation utilities, accelerating critical QA cycles and ensuring data integrity across various applications.
Education
SRM University Sikkim
Gangtok, Sikkim, India
→
Bachelor of Science
Information Technology
Skills
Cloud & Infrastructure
Google Cloud Platform (GCP), Microsoft Azure, Terraform.
Containers & Orchestration
Kubernetes, Docker, Helm, Kustomize, K3s, Slurm.
DevOps & Automation
GitLab CI/CD, GitHub Actions, Multi-stage Pipelines.
Monitoring & Security
Prometheus, Grafana, Loki, EFK Stack, Trivy, SonarQube, WAF, OWASP Rules.
Programming
GoLang, Python, JavaScript, Bash.
AI & HPC
CUDA, TensorRT, VLLM, SGLang, Kubeflow.
Projects
End-to-End MLOps Pipeline with Kubeflow
Summary
Built and maintained a Kubeflow-based pipeline enabling reproducible model training, validation, and deployment across GPU clusters.
Edge IoT Cluster Management System
Summary
Configured and managed K3s clusters for SmartCart IoT devices deployed in remote locations.
CI/CD and DevSecOps Automation Framework
Summary
Designed a multi-stage CI/CD pipeline (build, test, scan, deploy) utilizing GitLab CI and Terraform.