rootnomad@infra:~$ neofetch
_nnnn_ rootnomad@infra
dGGGGMMb ---------------
@p~qp~~qMb role : Systems & Infrastructure Engineer
M|@||@) M| os : Linux · FreeBSD (multi-boot brain)
@,----.JM| kernel : curious-6.x-hardened
JS^\__/ qKL shell : bash / zsh
dZP qKRb langs : Go · Rust · Python · C
dZP qKKb stack : bare metal -> control plane -> GPU/HPC
fZP SMMb editor : vim (and yes, I can exit it)
HZM MMMM uptime : building systems since first boot
FqM MMMM merged : 16 PRs upstream, latest first
__| ". |\dS"qML motto : make the failing layer obvious
| `. | `' \Zq
_) \.___.,| .'
\____ )MMMMMP| .'
`-' `--'I build and operate the systems underneath other systems — hardware, operating systems, networking, distributed software, cloud infrastructure, and production reliability — from first boot through automation, deployment, monitoring, and recovery. I like the faults that don't belong to one layer: a server healthy to its BMC that won't boot, a VM failing on a storage dependency, a GPU node with a driver or fabric problem, or a distributed op that has to recover safely after an interruption.
rootnomad@infra:~$ systemctl status rootnomad.service
● rootnomad.service — Systems & Infrastructure Engineer
Active: active (running) since first boot
CGroup: /infra.slice/rootnomad.service
├─ provisioning bare metal → first boot → ready pool
├─ reliability idempotent ops · health gates · safe rollback
├─ storage ceph · zfs · quorum · recovery paths
└─ gpu / hpc drivers · DCGM · InfiniBand · schedulingLanguages
Linux, cloud & containers
GPU / HPC / AI infrastructure
Bare metal & firmware
Storage & networking
IaC, CI/CD & observability
# 16 pull requests merged into major open-source infrastructure projects
apache/cloudstack · 12 merged
#14089reject creating a snapshot with a duplicate name for the same volume#14040engine: do not stall the host command queue on a failed send#14052server: treat a zero-capacity storage pool as full#14037server: scope IPv6 security group member rules to the exact host#14048server: report distinct read/write volume throttle rates… +7 more(#14050, #14047, #14027, #14025, #14024, #14023, #14022)
apache/cloudstack-terraform-provider · 3 merged
#339make network offering update work and read back correctly#338AND multiple filters in the pod and cluster data sources#337reject unsupported hypervisor values in host resource validation
ceph/ceph · 1 merged
#71496python-common: fixwith_units_to_int()crash on malformed size strings
| Area | What I work on |
|---|---|
| Systems & OS | Linux, FreeBSD, kernel/driver diagnostics, boot chains, OS images, install & recovery |
| Bare metal & hardware | BMC/Redfish/IPMI, firmware, inventory, PXE/UEFI, multivendor provisioning & lifecycle |
| Cloud & virtualization | Private cloud, KVM/QEMU/libvirt, Apache CloudStack, hypervisor networking, GPU virtualization, AWS |
| Distributed systems | Go & Rust services, control planes, durable state transitions, retry-safe operations |
| Containers & platforms | Kubernetes, k3s, node lifecycle, Docker/Podman, cluster operations & deployment automation |
| GPU / HPC / AI | Accelerator lifecycle, NVIDIA drivers/diagnostics, DCGM, Slurm, GPU scheduling, InfiniBand/RDMA/RoCE, PyTorch |
| Storage & data | Ceph, ZFS, block/file storage, RBD/CephFS, storage health, quorum, I/O & recovery |
| Networking | TCP/IP, routing, VLAN/VXLAN, bonding, DNS/DHCP/IPAM, load balancing, HA, debugging |
| SRE & automation | Production ops, on-call, telemetry, incident analysis, CI/CD, Terraform, Ansible, Jenkins |
| Security & trust | TLS/mTLS, PKI, OIDC, Secure Boot, FIPS-oriented hardening, signing, vuln remediation |
- Trace across layers. Hardware → firmware → kernel → network/storage → control plane → workload.
- Make failure modes explicit. Durable state, timeouts, rollback paths, a safe answer when uncertain.
- Automate with verification. Provisioning and releases are only done after meaningful health checks.
- Design for operators. Useful logs, dashboards, runbooks, reproducible fixes, repairable systems.
- Learn by building. Understand why it works, not just make it work once.
Most of my code is merged upstream (Apache CloudStack, Ceph, Terraform provider), so it lives in those orgs' histories rather than as stars on my own repos — see the merged PRs above for the real trail.
rootnomad@infra:~$ exit
logout
# thanks for scrolling. now go check your backups.Currently exploring: Linux & distributed-systems internals, GPU/HPC performance, PyTorch workload optimization, reliable control planes, and AI-assisted engineering workflows.
