Private infrastructure.
Open-source platform engineering.
Self-hosted LLM inference, Kubernetes at scale, and architecture for regulated industries where data cannot leave the premises.
View on GitHubPreviously worked with teams at
I build private AI platforms and production Kubernetes infrastructure. Self-hosted LLM inference, RAG pipelines, GitOps from bare metal to cloud. Background in fintech, pharma, and regulated industries.
More about how we work →AI infrastructure and open-source platforms.
Private AI Platform
Self-hosted LLM infrastructure that runs entirely on your premises. OpenAI-compatible API gateway, GPU inference on Kubernetes, RAG pipeline over your knowledge base.
Kubernetes Platform Engineering
Production-grade HA clusters on bare metal or cloud. Flux CD GitOps with encrypted secrets, Cilium eBPF networking, Prometheus/Grafana/Loki observability stack.
Regulated Industry Infrastructure
Architecture for regulated industries: air-gapped deployments, GDPR and MiFID II aligned, zero-trust network policy, secrets encrypted in git and decrypted only at runtime.
SRE & Incident Response
Senior SRE support: root cause analysis, post-mortem leadership, alerting gaps closed after every incident. AI-assisted incident diagnosis and automated GitOps remediation via Argus.
Private LLM Inference on Kubernetes: Real Costs and Real Numbers
An honest breakdown of what self-hosted LLM inference actually requires — hardware, costs, throughput reality, and when it makes sense versus just using the API.
Read → May 2026Secrets & Zero-Trust in Kubernetes
SOPS+age encrypted secrets, default-deny NetworkPolicies, and SSH hardening: complete secrets hygiene and zero-trust for a production Kubernetes cluster.
Read → May 2026Complete SRE Stack: Monitoring, Security & Disaster Recovery
kube-prometheus-stack, Falco runtime threat detection, and Velero: full observability, security, and disaster recovery layer for Kubernetes.
Read →Building on similar infrastructure?
Get in touch.
Short-term infrastructure engagements and platform work. [email protected] or LinkedIn.