NVIDIA-Certified Associate: AI Infrastructure and Operations
996e5a60-e96b-4a43-a37a-42d267a207f4
About Me
Senior AI Infrastructure Engineer | GPU/HPC Operations | LLM & AIOps | SRE | DevOps
Senior AI Infrastructure Engineer with 12+ years building and operating large-scale distributed systems, from enterprise cloud platforms (SUSE, Red Hat, Nokia) to GPU-scale AI infrastructure.
Currently operating a 4,000+ NVIDIA H100 GPU HPC cluster (500+ nodes, 8x H100 80GB HBM3 per node) supporting distributed LLM training, with full-stack ownership of production monitoring, self-healing alerting, auto-remediation, and LLM-powered operations tooling.
NVIDIA-certified in AI Infrastructure and Generative AI LLMs. Hands-on with GPU cluster management, CUDA/NCCL workloads, InfiniBand networking, and production LLM agent development with RAG, MCP, and multi-model architectures.
AI / LLM Ops
Building LLM-powered operations agents with RAG, MCP, and NL-to-SQL. Multi-model support (Claude, OpenAI, Ollama) for infrastructure diagnostics.
GPU / HPC
Operating a 4,000+ NVIDIA H100 GPU cluster (500+ nodes). DCGM, CUDA/NCCL, SLURM scheduling, InfiniBand fabric, GPUDirect RDMA, and distributed training orchestration.
SRE / Platform
Self-healing monitoring with 12+ background modules, auto-remediation workflows, alert deduplication, and unified observability (Prometheus, Grafana, TimescaleDB).
Cloud / DevOps
Infrastructure as Code, CI/CD pipelines, container orchestration, and multi-cloud architecture across AWS, Azure, GCP, OpenStack, and OCI.
Skills
AI / LLM
- LLM Agent Development
- RAG (pgvector)
- Model Context Protocol (MCP)
- Prompt Engineering
- Claude / OpenAI / Ollama APIs
- NL-to-SQL Compilation
- AIOps & AI-Driven Automation
GPU / HPC
- NVIDIA DCGM
- CUDA / NCCL
- NVLink Diagnostics
- SLURM
- InfiniBand / RDMA / RoCE
- OpenMPI
- HPC Benchmarking (IOR, HPL)
ML Training
- PyTorch / torchrun
- FSDP (Distributed Training)
- Mixture of Experts (MoE)
- MosaicML Composer
- Flash-Attention
- Weights & Biases / TensorBoard
- Training Orchestration (4,000+ GPUs)
Observability
- Prometheus / Telegraf
- Grafana
- TimescaleDB
- Custom Monitor Stacks
- Slack-Integrated Alerting
Languages
- Python
- Bash
- Nix
- YAML
Cloud
- AWS
- Azure
- GCP
- Oracle Cloud (OCI)
- OpenStack
- Ceph Storage
IaC / DevOps
- Terraform
- Ansible
- Chef / SaltStack
- Git / CI/CD
- Jenkins / Hydra
Containers
- Docker
- Kubernetes (CKA)
- NVIDIA Container Runtime
- Docker Compose
Networking
- InfiniBand (ConnectX-6)
- Mellanox UFM
- Dell SONiC
- SNMP / RDMA / UCX
- Network Automation
Data
- PostgreSQL / TimescaleDB
- Redis
- Elasticsearch
- Pandas / NumPy
Hardware / DC
- Dell iDRAC / Redfish
- IPMI / BIOS (racadm)
- VAST Data / NFS
- PSU / Thermal Monitoring
- 500+ Node Cluster Ops
Frameworks
- FastAPI / Uvicorn
- Slack Bolt SDK
- Chainlit
- REST / Redfish APIs
Linux
- RPM / DEB / NixOS
- Server Hardening
- Kernel Tuning
- SSH Orchestration
Certifications
NVIDIA-Certified Associate: Generative AI LLMs
a4647c10-d4bf-4dde-9452-514413e6e291
Certified: Terraform Associate (003)
aeceb2f3-33ac-4ad4-a94a-dbb177fc619a
Certified Kubernetes Administrator
LF-a2o0xdlc96
Certified OpenStack Administrator
COA-1600-0067-0100
AWS Certified Cloud Practitioner
K40MK0MDFEEQ13CJ
OpenStack Administrator Certification Professional Level
200-422-511
Red Hat Certified Administrator
150-132-256
Linux Foundation Certified Engineer
LFCE-1500-0058-0200
Certified Ethical Hacker v9
ECC55284410087
ITIL Foundation Certificate in IT Service Management
5426913.20427512
IBM Certified System Administrator - AIX 7
03005008
SUSE Enterprise Architect
10186201
SUSE Certified Instructor
10186201
Certified Engineer Enterprise Linux
10186201
Certified Engineer OpenStack Cloud
10186201
Certified Administrator Enterprise Storage (Ceph)
10186201
Certified Administrator Systems Management (SUSE Manager)
10186201
Projects
SlaSH
Slack bot for GPU datacenter operations. Real-time GPU health monitoring (DCGM, NVLink, ECC), InfiniBand per-lane error analysis, and 13-test Dell diagnostics bundle with automated TSR collection. Background monitors for thermal, network, and NFS health with auto-remediation and intelligent alert deduplication.
Elliot
LLM-powered operations agent for GPU datacenter diagnostics. Natural language to SQL against TimescaleDB, protocol-aware discovery (Redfish/SNMP/SSH), and tool evolution from repeated query patterns. Multi-interface: Slack, REST API, CLI, and Chainlit web UI.
PapayaPlaya
LLM training orchestration for GPU clusters without Slurm dependency. Manages NGC container distribution, NFS-backed checkpoint storage, and torchrun-based distributed training. Intelligent image transfer logic and automated training result analysis with W&B integration.
sysBOFH
Personal flake-based NixOS configuration system supporting reproducible setups across NixOS desktops, servers, and macOS machines using shared, modular components.
DevOps Nix Shell
Portable, reproducible nix shell environments with all your DevOps tools and dotfiles, run them anywhere, instantly.
BOFH NixVim
Functional and minimalist NixVim configuration, featuring a carefully curated set of essential plugins for an efficient, distraction-free editing experience.
Anvil
Containerized GPU cluster benchmarking platform. Orchestrates NCCL collective tests, HPL compute benchmarks, and DCGM thermal profiling across distributed nodes with multi-node SSH orchestration and Prometheus-based metric collection.
5cents-cluster
Containerized NCCL test runner for single and multi-node GPU configurations. Tests collective communication performance (all_reduce, all_gather, broadcast) with dual build modes and automated node connectivity validation.
IoRi
Distributed I/O benchmarking tool for GPU clusters. Automated containerized IoR tests with OpenMPI across multiple nodes, flexible node range specification, and pre-flight connectivity validation with structured result collection.
OpenStack Lab
Zero-touch OpenStack deployment pipeline: Terraform provisions the infrastructure via libvirt, cloud-init bootstraps the OS, and Ansible drives DevStack end-to-end. One command from bare metal to a fully operational private cloud with Keystone, Nova, Neutron, and Horizon.
Ansible Lab
Dynamically scalable Ansible sandbox built on Docker Compose. Spin up N target nodes on demand with a single parameter, all SSH-ready with auto-provisioned keys. Custom openSUSE Tumbleweed image stripped of systemd for a minimal, production-realistic footprint.
CEPH Lab
Scalable Ceph distributed storage cluster with isolated dual-network topology (admin + cluster replication). Terraform provisions MON, OSD, and admin nodes on demand; Ansible drives ceph-deploy orchestration with Vault-managed secrets. From zero to a production-grade storage fabric in minutes.
Get In Touch
Feel free to reach out through any of these channels.