NestJS & Kubernetes consultant · Founder at Totem Labs Forge

Not a vibe coder. A senior engineer with the architecture background AI needs to actually ship something real.

vikgmdev.com · totemlabs.io · eraastrology.ai · forgetty.dev

_{✨  Now shipping  ✨}

ERA — Cosmobiology AI
_{eraastrology.ai · three-generation lineage, live on Android}

_{Ship-to-Live → you vibe-coded a working app. In 10 business days it's deployed, monitored, backed up, and alerting you when it breaks. $497 Audit · $1,500 Core (done-for-you in 10 days) · $2,500 Launch (5-day rush + launch-weekend war room). Free playbook → github.com/vikgmdev/ship-to-live}

You vibe-coded an app. You need it in production. → Ship-to-Live: I audit it, containerize it, deploy it to Kubernetes (mine or yours), wire monitoring + backups + CI/CD, and stay on call for 30 days.

$497 Audit · $1,500 Core (done-for-you in 10 days) · $2,500 🚨 Launch (5-day rush + launch-weekend war room) · $500/mo ongoing ops

Free self-serve playbook → github.com/vikgmdev/ship-to-live

A decade as a Backend Engineer first. Node.js / TypeScript / NestJS / GraphQL / microservices — before anything else, that's the root. On top of that: years of hands-on infrastructure at bare-metal scale (Kubernetes, Ansible, Kafka, Postgres HA, Elasticsearch, Redis Cluster), and the last ~2 years going deep on AI — building MCP servers in production, running multi-agent harnesses, and shipping AI-native products end-to-end.

I run two parallel tracks today:

1. Senior Backend / DevOps at a blockchain security platform — platform, API, observability, on-call.
2. Founder & sole engineer at Totem Labs Forge LLC — a Delaware product studio where I design, build and ship AI-first products end-to-end under a custom multi-agent harness.

AI coding tools are turning a lot of people into "vibe coders" — shipping surface-level demos that fall apart the moment they hit real load, real state, or a real incident. I'm the opposite. I came up the hard way: wire protocols, daemon architectures, stateful migrations, SCRAM auth, pgBackRest DR, CI/CD pipelines written by hand.

Because I already know how systems work at the byte level — how a Kafka broker negotiates TLS, how a Cilium pod masquerades through a vSwitch, how a Vault AppRole round-trips a token, how @elastic/elasticsearch shape-drifts between v7 and v9 — I can let AI do the drudge typing while I keep ownership of the architecture. The result is not a demo. It's ~38K lines of production Rust in 24 days (Forgetty), an AI app live on Google Play (ERA) on a cost-controlled three-tier model stack, a production MCP server, and a harness-driven Kubernetes cluster that a teammate can re-stand-up in an afternoon.

AI is the multiplier. Backend + infra engineering is the foundation that lets the multiplier do anything useful.

This is my strongest surface. I don't "know" Kubernetes, Redis, Postgres and Elasticsearch from a course — I operate all four in production, on real traffic, with real users downstream, at the same time.

The common thread: I know these four at the byte level — how a Redis protected-mode handshake rejects, how pgBackRest asyncs a full backup to SFTP, how Cilium masquerades pod source IPs through a vSwitch, how @elastic/elasticsearch drifts between v7 and v9. That's what lets me ship zero-regression migrations instead of blog posts.

Live on Google Play since April 2026. Built on a methodology transmitted directly from Hector García — a master astrologer with 30+ years of practice, my father — so ERA reasons from a real interpretive lineage rather than "trained on data".

MIT on github.com/vikgmdev/forgetty. Daemon-first architecture (PTY + byte-log + session persistence in the daemon; renderers own VT parsing via libghostty-vt FFI). Binary length-prefixed wire protocol. Tabs + split panes, per-pane scrollback, search, 486 bundled themes, command palette, session restore, named workspaces, socket API. 24 days / ~38K LOC of production Rust despite no prior Rust experience — evidence that the Forge harness works: it handles the drudge typing while I own the architecture decisions.

The methodology that made Forgetty and ERA possible at solo-founder velocity. A three-phase subagent-isolated workflow:

• Plan — spec + acceptance criteria agent
• Build — implementation agent with mandatory simplify / audit-rust / audit-security passes
• QA — human-in-the-loop QA scored 1–10 on five axes, binary pass/fail

Per-task git worktrees let me run 3–5 feature sessions in parallel; overnight autonomous mode completes tasks unattended with deferred QA for the morning.

_{Operator-grade experience across Cosmos · Stellar · XRPL EVM. Solana: RPC integrations and on-chain indexing (not pitched as a protocol specialist). Solidity smart-contract exposure.}

Short, paid engagements for teams that want real engineering on top of AI — not the other way around.

You vibe-coded an app. You need it in production. → Ship-to-Live: I audit it, containerize it, deploy it to Kubernetes (mine or yours), wire monitoring + backups + CI/CD, and stay on call for 30 days.

$497 Audit · $1,500 Core (done-for-you in 10 days) · $2,500 🚨 Launch (5-day rush + launch-weekend war room) · $500/mo ongoing ops

Free self-serve playbook → github.com/vikgmdev/ship-to-live

• NestJS expert — migrations to NestJS from Express / Koa / vanilla Node / monolithic TS backends. Architecture, module boundaries, DI done right, GraphQL / REST / WebSockets / MCP resolvers, testing and CI/CD. Shipped large NestJS codebases in production — I know how to land the migration without freezing your feature work.
• Kubernetes on bare metal — bootstrap from scratch, reproducible in an afternoon. kubeadm HA control plane, Cilium eBPF, ArgoCD GitOps, External Secrets Operator + Vault, Cloudflared Zero Trust ingress, observability that actually pages the right person.
• Elasticsearch migrations — major-version upgrades on live platforms. Shape-compare safety net, dual-write phases, per-service SPEC → Build → QA, reversible flip-back. Zero regression is the target, not a wish.
• PostgreSQL HA + real DR — streaming replication, pgBackRest to off-provider storage, automated restore testing so you find out DR is broken on Tuesday afternoon, not at 3am.
• Redis — Cluster + hardening — multi-instance topology per workload, end-to-end lockdown (VLAN + localhost + protected-mode + auth), Prometheus federation + paging. No more caches listening on 0.0.0.0.
• Kafka — TLS + SASL/SCRAM + per-consumer ACLs, the streaming backbone without the footguns.
• AI-multiplied engineering for your team — how to plug Claude Code / MCP / multi-agent harnesses into your real SDLC without turning your codebase into a vibe-coded graveyard. Architecture-first, shipping-focused.

• vikgmdev.com — portfolio with technical case studies
• Kubernetes from scratch on bare metal — 10-script reproducible bootstrap
• Elasticsearch v7 → v9 migration on a live platform — shape-compare strategy, zero regression

_{Profile scaffolding inspired by the open-source career-ops and cv-santiago by @santifer. Thank you for the blueprint 🙏}