1
yana-stocks: A Next.js Microservices Platform
An 11-service polyglot platform for real-time stock data, portfolios and ML predictions — service boundaries, layered caching, tracing, and change-aware CI.
Blog
Writing about things I'm building, learning, and tinkering with — from production engineering to home servers.
SeriesApplications
Deep dives into the full-stack applications running on the on-prem cloud — architecture, backend design, APIs, frontend performance, and observability, end to end.
1 post
Series
Applications
Deep dives into the full-stack applications running on the on-prem cloud — architecture, backend design, APIs, frontend performance, and observability, end to end.
1 post
SeriesRAG AI
Building retrieval-augmented generation systems — how RAG works, and what building one over the on-prem cloud’s own docs reveals about the approach.
1 post
Series
RAG AI
Building retrieval-augmented generation systems — how RAG works, and what building one over the on-prem cloud’s own docs reveals about the approach.
1 post
1
A RAG Chatbot for the On-Prem Cloud’s Docs
What retrieval-augmented generation actually is, and how a RAG chatbot over k8s-apps’ docs was built with OpenAI embeddings and Claude for cited answers.
SeriesAgentic AI
Building production agentic systems with deepagents and LangGraph — orchestrator/subagent architectures, human-in-the-loop remediation, and what actually breaks when an agent gets to act, not just answer.
1 post
Series
Agentic AI
Building production agentic systems with deepagents and LangGraph — orchestrator/subagent architectures, human-in-the-loop remediation, and what actually breaks when an agent gets to act, not just answer.
1 post
1
ops-agent: A LangGraph Agentic SRE Copilot
How ops-agent’s orchestrator delegates to fifteen scoped subagents, and the LangGraph mechanics — interrupt/resume, a CNPG-backed checkpointer — behind it.
SeriesBuilding a Production On-Prem Cloud
An 8-part series on setting up a production-grade Kubernetes cluster at home — storage, networking, databases, GitOps, SSO, and shared application services. See the stack — Grafana, Gotify, Hubble, Uptime Kuma, Harbor, MinIO, Kafka, MongoDB, Redis, Nextcloud, Immich →
8 posts
Series
Building a Production On-Prem Cloud
An 8-part series on setting up a production-grade Kubernetes cluster at home — storage, networking, databases, GitOps, SSO, and shared application services. See the stack — Grafana, Gotify, Hubble, Uptime Kuma, Harbor, MinIO, Kafka, MongoDB, Redis, Nextcloud, Immich →
8 posts
1
Building a 6-Node Kubernetes On-Prem Cloud
What I learned building a production Kubernetes cluster at home — Proxmox, Cilium native routing, Ceph CSI, ArgoCD, and the gotchas not in the docs.
2
Ceph CSI in Kubernetes: Bare-Metal Storage
Connecting a Ceph cluster to Kubernetes via the CSI driver for dynamic RBD provisioning — including the Cilium native routing gotcha that blocks OSD traffic.
3
Cilium Network Policy in Native Routing Mode
Three scenarios where standard NetworkPolicy breaks down in Cilium native routing mode, and how CiliumNetworkPolicy’s toEndpoints and toCIDR solve them.
4
CloudNativePG: Running Postgres in Kubernetes
Using CloudNativePG for database-per-service Postgres — shared vs dedicated clusters, barman WAL backups to MinIO, and the kured drain gotcha.
5
ArgoCD: GitOps for an On-Prem Cloud That Takes Itself Seriously
Sync waves, ignoreDifferences, the Argo Rollouts diff gotcha, permanent OutOfSync apps, and the valuesObject pitfall — ArgoCD on real workloads.
6
Authentik: SSO for Your On-Prem Cloud
Putting every on-prem cloud service behind single sign-on with Authentik — the forward auth pattern, the auth-url hostname gotcha, and native OIDC for Grafana.
7
shared-services: A Kafka-Mediated Email Pipeline (And Whether It Was Worth It)
Replacing three duplicated SMTP2GO integrations with one Kafka-mediated email service — the Harbor robot account gotcha, and whether it was over-engineered.
8
From Flannel to Cilium: A Live CNI Migration
The cluster didn’t start on Cilium. Why Flannel’s silently-unenforced NetworkPolicies forced a migration, and what eBPF native routing and Hubble buy you.