Projects
What I build outside work, mostly to learn infrastructure end to end: from a Kubernetes platform with its own MLOps stack to the servers that run my home.
MaetisHUB
A cloud-native, multi-purpose AI platform I build and run on Kubernetes, from 2025 to today. Private repository, 1,000+ commits.
-
Platform
GitOps delivery with Argo CD and canary rollouts, an Istio service mesh with mTLS, secrets in Vault, OPA Gatekeeper policies and Keycloak for identity.
-
Data and ML
PostgreSQL with pgvector, an S3-compatible lakehouse on MinIO, Redpanda streaming, Dagster pipelines with OpenLineage lineage, a Feast feature store, MLflow and KServe model serving.
-
AI services
An LLM gateway and an autonomous agent with sessions, tools and audit trails, plus computer vision and NLP domains.
-
Observability
Metrics, logs and traces with Prometheus, Grafana, Loki and Tempo, so every service is measurable from day one.
LLM assistant
A personal assistant I built: retrieval-augmented generation with reranking and a two-pass draft-and-verify pipeline on LangGraph, reachable from Telegram, Alexa and the web.
Home lab
Three machines on a private WireGuard mesh, with a repurposed laptop running as a 24/7 Docker server. Nothing is exposed to the public internet.
-
Home automation and security
Home Assistant for automation and presence, and a network video recorder with on-device object detection on the GPU.
-
Network
Mesh networking with Tailscale, network-wide DNS filtering, and dashboards served only inside the mesh.
-
Data safety
Continuous file sync with versioned copies across machines, and encrypted off-site backups.
This site
Plain HTML and CSS with a little progressive JavaScript, served from Cloudflare's edge. Every word is in the HTML, so search engines and AI assistants can read it without running scripts.