Running from my home, reachable worldwide
A two-node homelab where I design, deploy, observe, and operate real applications—exposed through a VPS without opening my home network directly to the internet.
# deliberately small, intentionally real
platform:
nodes: [compute, storage]
ingress: VPS → WireGuard
runtime: Docker + Proxmox
access: OAuth + private mesh
signals: metrics, logs, health
● all public paths operational
Software Engineer III
About me
I’m Anikesh, a Software Engineer III with 4+ years of experience building reliable backend and full-stack systems. At work, I focus on distributed services, data platforms, performance, and compliance. At home, I run the entire lifecycle—from an idea and a pull request to networking, deployment, observability, and recovery.
Architecture explorer
This is not a cloud diagram assembled for a slide. Select a path to see how the running platform handles traffic, data, and operations.
The VPS is the only public edge. Traffic enters the home network through an encrypted tunnel and terminates at the intended application.
Health probes, route metrics, container signals, and alerts make this path diagnosable.
Selected systems
Personal projects with production concerns: authentication, queues, caching, safe automation, deployment, monitoring, and recovery.
Application platform
A single authenticated surface for the tools I actually use: price tracking, coding practice, flight monitoring, fleet diagnostics, and long-running workflows.
Distributed application
Scheduled product checks across stores with durable state, browser-based extraction, a reactive API, history, and user-facing alerts.
Developer tooling
A self-hosted web console over coding agents, with separate general, tool-building, and operator workflows plus allowlisted workspace access.
Networking & security
A public Caddy edge with TLS and route controls, joined to the homelab over WireGuard so the residential network is never the public endpoint.
Operations
Health probes, host and container metrics, route analytics, real-time SSE diagnostics, and alerts spanning the compute and storage nodes.
Private cloud
Private photo backup, file storage, and controlled request-to-playback workflows, with compute-heavy services separated from bulk storage.
Purva Labs / open source
I use Purva Labs to turn the useful parts of my homelab into focused, documented tools that other people can run themselves.
A self-hosted interface for coding agents with persistent conversations, General, Tool Builder, and allowlisted Operator modes.
purva-labs/agent-console ↗A focused dashboard for container status, resource usage, logs, lifecycle actions, safe rebuilds, and rollback on recreation failure.
purva-labs/docker-control ↗A lightweight web file manager scoped to one configured root, with previews, uploads, search, and protection against path escapes.
purva-labs/file-manager ↗Engineering decisions
A homelab is useful precisely because resources are finite and failure is personal. These choices keep the system understandable and recoverable.
DECISION / STORAGE
Latency-sensitive databases and configuration live on the compute node's local disk. The Raspberry Pi remains the authoritative NFS storage node for photos, media, and backups.
The storage node can still affect dependent apps, so mounts and startup order are monitored explicitly.
Four years into building production software
This platform is where product engineering, distributed systems, infrastructure, and operational discipline meet.