I'm a Senior Linux & Network Engineer who designs, builds and runs secure, scalable, high-performance infrastructure for research institutes, universities and government organizations across the Netherlands, Belgium and Germany.
From the datacenter fabric to the GPU, I work across the full stack. That means EVPN/VXLAN leaf-spine networks, hardened Linux platforms, virtualization and containers, identity and access management, and the automation that holds it all together.
At CWI, the National Research Institute for Mathematics and Computer Science, I have the freedom and resources to build experimental environments at real scale. I use them to test new network architectures, put emerging technologies under load, and turn research ideas into infrastructure that works in production.
Right now, I'm focused on NVIDIA DGX platforms, GPU computing, and local AI, including combining NVIDIA DGX Spark with Exo to build decentralized, privacy-first AI infrastructure that runs entirely on-premises.
And when life gets too predictable, I break my homelab on purpose to learn how to rebuild it better. π§
| π Building | Sovereign, self-hosted platforms for research & government |
| π§ Exploring | DGX Spark + Exo for distributed, privacy-first local AI |
| π Researching | Post-quantum cryptography: hybrid SSH key exchange, OpenPGP & age |
| π Fabric-ing | EVPN/VXLAN & RoCEv2 |
| πΏ Living by | The Joy of Missing Out (JOMO) |
|
Fabrics that don't fall over. |
From kernel to cluster. |
|
If it's done twice, it's code. |
Zero Trust by design, not as an afterthought. |
|
GPUs need networks that keep up. |
Prototype at scale, then make it production-grade. |
π Networking & Routing
Ethernet IPv4 IPv6 VLAN 802.1Q QinQ LACP LAG MLAG MC-LAG STP RSTP MSTP LLDP ARP NDP OSPF OSPFv3 BGP iBGP eBGP MP-BGP BFD ECMP IS-IS VRF PBR RIB FIB ASN RPKI Anycast Multicast IGMP PIM
π’ Datacenter Networking & SDN
EVPN VXLAN VNI VTEP BGP EVPN IRB Symmetric IRB Anycast Gateway Route Distinguisher Route Target ARP Suppression BUM GENEVE MPLS SR-MPLS SRv6 Leaf-Spine Clos Fabric SONiC NVIDIA Spectrum Mellanox
π Network Security & Encryption
MACsec MKA 802.1X EAP IPsec IKEv2 WireGuard OpenVPN TLS mTLS PKI X.509 ACL NGFW IDS IPS DPI RADIUS TACACS+ NAC ZTNA Zero Trust SIEM SOC MFA FIDO2 YubiKey HSM TPM
π§ Linux & Systems Engineering
Linux Ubuntu Debian systemd GRUB UEFI Secure Boot DKMS cgroups Namespaces SELinux AppArmor PAM POSIX glibc D-Bus udev eBPF io_uring NUMA Bash Python SSH GPG
πΎ Virtualization, Storage & Distributed Systems
KVM QEMU Proxmox VE VMware ESXi Xen libvirt LXC Incus SR-IOV PCI Passthrough VFIO VirtIO Open vSwitch DPDK Live Migration HA LVM RAID ZFS Btrfs ext4 XFS NFS SMB iSCSI NVMe NVMe-oF Ceph GlusterFS Lustre Object Storage S3 Snapshots Deduplication
π¦ Containers & Orchestration
Docker Podman containerd CRI-O OCI Kubernetes K3s Helm Kustomize CNI CSI Calico Cilium Flannel MetalLB Istio Envoy Traefik NGINX HAProxy Service Mesh Gateway API
βοΈ Automation, DevOps & Infrastructure
Puppet SaltStack Git Forgejo GitLab GitHub Actions Jenkins CI/CD GitOps Argo CD Flux CD Packer cloud-init PXE iPXE ZTP REST API gRPC NETCONF RESTCONF YANG Jinja2 YAML JSON
π NVIDIA, GPU Computing & HPC
NVIDIA DGX DGX Spark Mellanox Networking NVIDIA Spectrum InfiniBand RDMA RoCEv2 NVLink CUDA NVIDIA GPU Drivers NVIDIA Container Toolkit GPU Passthrough GPU Virtualization
π§ Artificial Intelligence & Machine Learning
Local AI Self-Hosted AI LLMs Deep Learning NVIDIA CUDA PyTorch Ollama llama.cpp vLLM Exo Hugging Face Docker AI Containers RAG Model Quantization Distributed Inference
π Monitoring, Observability & Operations
Zabbix Grafana NetBox Syslog rsyslog NetFlow sFlow
Note
In a single decade, a handful of hyperscalers (Google, Microsoft and Amazon Web Services) have come to host a staggering share of the world's digital infrastructure and data.
Governments, research institutions and universities are waking up to what that costs them: sovereignty, privacy, freedom from vendor lock-in, and long-term independence.
I build the alternative: open-source, self-hosted and fully under your own control.
| π€ Collaboration Self-hosted platforms |
βοΈ Secure mail infrastructure |
πͺͺ Identity IAM & SSO systems |
π Networks Modern Linux & network architecture |
π Services Open, transparent enterprise IT |
Quantum computing and ever more powerful AI mean today's cryptography has an expiry date. I don't see that as a threat to manage. I see it as a rare chance to rebuild security on stronger foundations.
The Quantum Threat Timeline Report 2023 maps out when quantum computers may be able to break today's public-key cryptography (see the expert-opinion chart).
Important
The post-quantum era is no longer theoretical. The transition has already begun. Data harvested today can be decrypted tomorrow.
My focus is real-world usability, not theory. I'm running hybrid post-quantum key exchange for SSH in practice:
| Mechanism | Spec |
|---|---|
| π ML-KEM hybrid key exchange for SSH | draft-ietf-sshm-mlkem-hybrid-kex |
| π NTRU Prime for SSH | draft-josefsson-ntruprime-ssh |
I've also put OpenPGP and age through their post-quantum paces. OpenPGP is powerful, but its PQ transition is slow and complex. The OpenPGP PQC draft is promising, yet concepts such as Persistent Symmetric Keys (PSKs) remain hard to reason about, both for users and for operators. age, by contrast, has a clean and understandable model that already makes a solid foundation for post-quantum workflows.
βοΈ OpenPGP vs age: a practical comparison
| ποΈ OpenPGP | β¨ age |
|---|---|
| Mature ecosystem | Modern and minimalist |
| Complex implementation details | Easy to understand and experiment with |
| Slower PQ transition path | Clean and flexible design |
| Fragmented tooling | Simple, consistent workflows |
| Powerful but hard to operationalize | Practical and approachable |
Tip
Verdict: while both ecosystems are still evolving, age is leaner, clearer, and easier to build modern post-quantum workflows around.
Open standards, technical independence, security and relentless experimentation are the foundations of infrastructure that lasts.
I combine proven engineering discipline with a willingness to challenge the default, so the systems I build are reliable today and ready for whatever tomorrow demands. Most of all, they stay understandable and maintainable long after the first deploy.
benjamin@infra:~$ cat /etc/philosophy.conf
[principles]
β Keep systems understandable
β Prefer open over opaque
β Automate the boring parts
β Break things β learn β rebuild better
β Own your data, or someone else will
benjamin@infra:~$ systemctl status curiosity
β curiosity.service - Always building, always learning
Loaded: loaded (/etc/systemd/system/curiosity.service; enabled)
Active: active (running) since forever