On-Prem LLM on DGX Blackwell (GB10)
Fine-tuned Qwen & Gemma for client domains, applied NVFP4 / Gemma4 quantization for constrained GPU memory, and served via Ollama & vLLM on an NVIDIA DGX Blackwell (GB10) production Kubernetes cluster.
AWS Certified DevOps Engineer — Professional · Bengaluru, open to remote
Senior DevOps / MLOps / AI Infrastructure Engineer with 7+ years of experience. I design, automate and secure the Kubernetes, GPU/LLM serving and CI/CD platforms behind machine-learning workloads — across AWS, GCP and Azure.
Selected work
Production-shaped work from client deliveries and freelance engagements — not tutorials.
Fine-tuned Qwen & Gemma for client domains, applied NVFP4 / Gemma4 quantization for constrained GPU memory, and served via Ollama & vLLM on an NVIDIA DGX Blackwell (GB10) production Kubernetes cluster.
Containerized real-time CCTV analytics for a large airport: HAProxy fronting UI, REST API, Kafka and MongoDB, with DMZ segmentation and security-zone isolation.
Modular Terraform (VPC / IAM / EKS CPU+GPU) with Ansible hardening; MLflow, DVC, Prometheus, Grafana and ClickHouse, secured with HashiCorp Vault and RBAC.
Real-time IoT pipeline — GCS → Cloud Run → BigQuery — built with Terraform and GitHub Actions, cutting misconfigurations by 70% and manual effort by 90%.