Operator-First Infrastructure & AI Systems

Engineered reliability, not résumé copy.

I design and run infrastructure that stays understandable under load: self-hosted platforms, multimodal RAG copilots, ingress, DNS, observability pipelines, and cloud foundations with clear change paths. Every component here represents real operational work, not placeholder projects.

Explore Projects Read Medium Articles
About & Impact

Standardizing foundations, automating change, proving patterns.

Current Role

DevOps Engineer, HashedIn by Deloitte

Problem Keep cloud and platform delivery repeatable, secure, and easy to hand off without tribal knowledge.
Solution Cloud infrastructure, Kubernetes, automation runbooks, and documentation-first delivery.
Impact Cleaner change flow, automated CI/CD pipelines, and reduced operational complexity.
Independent Practice

Homelab & AI Platform Engineering

Problem Self-hosted services fail fast when ingress, DNS, backups, telemetry, and vector retrieval drift.
Solution Qdrant vector search, Frigate object detection, Traefik, OPNsense NAT routing, and staged automation.
Impact A battle-tested proving ground for production-grade network and AI infrastructure.
Architecture Case Studies

The homelab as one cohesive system.

Network flow diagram showing OPNsense, AdGuard, Twingate, Tailscale, Nginx reverse proxy, and Docker stack.
Ingress & Remote Access

Network Flow & Ingress Topology

Problem Remote access had to stay clear and debuggable across local browsing, Tailscale VPN, and public ingress.
Stack OPNsense, Unbound, AdGuard, Traefik, Nginx, Twingate, Tailscale, Docker.
Architecture Traffic planes are split so DNS resolution, app reverse proxying, and remote access operate independently.
Result Deterministic troubleshooting, zero subnet overlap conflicts, and clean ingress routing.
GitHub Repo Medium Article
Service topology diagram showing Proxmox hosts, VMs, LXCs, and Docker workloads.
Workload Placement

Hosts, VMs, LXCs & Docker Workloads

Problem A multi-node homelab becomes chaotic unless control plane, utility services, and media apps are cleanly grouped.
Stack Proxmox VE, Ubuntu Server, Docker, Home Assistant, Jellyfin, Nextcloud, Grafana Loki.
Architecture Isolate edge ingress, telemetry, storage, and media workloads so hardware failure domains stay isolated.
Result Transparent placement decisions, explicit resource allocation, and zero host-level sprawl.
GitHub Repo Medium Article
Virtual networking diagram showing Proxmox bridges, TAP/VETH pairs, and Docker custom bridges.
L2 / L3 Boundaries

Bridges, NAT & Container Networking

Problem Physical NICs, Proxmox bridges, and container namespaces can blur together without an explicit layer model.
Stack Proxmox vmbr0, TAP devices, VETH pairs, Docker custom bridge, netfilter/iptables, Linux namespaces.
Architecture Map the path from physical NIC to bridge to workload namespace so L2 bridging and L3 routing stay separate.
Result Clear packet traversal diagrams for network diagnostics and secure VLAN trunking.
GitHub Repo
Featured Projects

Production systems & engineering proofs.

AI / Vector Retrieval

Homelab Copilot: Self-Hosted Multimodal RAG

Status: Active

Built a 100% self-hosted AI assistant indexing scattered homelab markdown notes, Draw.io XML graph topologies, Excel IP maps, and live Grafana Loki container log streams into a unified hybrid vector search engine.

Problem Searching through 50+ markdown runbooks, spreadsheet IP tables, Draw.io diagrams, and Loki log streams to diagnose container/network failures was slow and fragmented.
Stack Qdrant Vector DB, FastEmbed BM25, MiniLM-L6-v2, Cross-Encoder (ms-marco-MiniLM-L-6-v2), Python, Grafana Loki API, openpyxl, zlib.
Architecture Header-aware markdown chunking, base64 XML graph extraction for Draw.io diagrams, dual-tier adaptive Loki log chunking, Qdrant Reciprocal Rank Fusion (RRF), and cross-encoder logit thresholding (-2.0 cutoff).
Impact Achieved 100% Hit@1 accuracy and 1.0 MRR on a 17-item homelab test suite with sub-700ms retrieval and zero cloud data leakage.
GitHub Code Deep-Dive Article
Networking

Six-VLAN Homelab Segmentation

Status: Active

Segmented a flat home network into six purpose-built VLANs with L3 routing and mixed VLAN-aware/unaware hardware.

Problem A flat network mixes management, IoT, security cameras, and servers into one broadcast domain and blast radius.
Stack GWN7721 managed switch, OPNsense (L3 routing, mDNS repeater), VLAN trunking.
Architecture Six isolated VLANs (management, trusted, servers, cameras/NVR, IoT, guest); native VLAN trunking allows non-VLAN APs to deliver untagged traffic cleanly.
Impact Cameras, IoT devices, and core servers sit in strict trust domains with explicit inter-VLAN firewall rules.
GitHub Repo
Edge AI / Media

Frigate NVR on Constrained Hardware

Status: Active

Tuned a multi-camera Frigate NVR deployment to run real-time object detection on a repurposed consumer laptop GPU.

Problem Real-time object detection across 6+ camera streams overwhelms consumer CPU/GPU hardware without stream tuning.
Stack Frigate NVR, ONNX/GPU detector, go2rtc, ffmpeg, Acer Aspire 7 (GTX 1650 4GB, Ryzen 5).
Architecture Dual go2rtc streams per camera (high-res record, low-res detect at 640x360 5fps); hardware-accelerated decoding and tuned zone alert rules.
Impact Stable multi-camera real-time detection and identity-aware alerting on budget hardware.
GitHub Repo
Cloud Platform

Azure Landing Zone Foundations

Status: Supporting

Module-driven Azure platform example demonstrating IaC structure, peering, and landing zone principles.

Problem Demonstrate enterprise cloud landing zone architecture in clean, public-safe Terraform code.
Stack Azure, Terraform, reusable modules, VNet peering, architecture docs.
Impact Provides a clean reference implementation for repeatable Azure cloud foundations.
GitHub Repo Medium Article
Developer Tooling

Hybrid Local + Remote AI Pipeline

Status: Active

Hybrid local/remote LLM coding setup built to operate smoothly across large multi-repository codebases.

Problem Single AI providers hit memory limits, context bottlenecks, or rate limits during multi-file coding sessions.
Stack mlx-lm (Apple Silicon), GitNexus (MCP), Aider, Continue.dev, Qwen 2.5 Coder 14B, Groq/OpenRouter fallbacks.
Impact Resilient coding workflow with automated prompt-refinement agents and seamless mid-session model switching.
GitHub Profile
Technical Articles

Curated engineering writeups from Medium.

Networking & Ingress

Remote Access / Tailscale

Connecting Remote IP Cameras to Frigate via Tailscale: A Network Guide

02 Oct 2026 · ~8 min read

Debugging asymmetric routing, same-LAN gateway bypass, and TCP handshake failures across site-to-site Tailscale subnet routers using OPNsense Outbound NAT.

Read Article on Medium
Ingress / Reverse Proxy

Traefik Done Properly in a Homelab: Nextcloud, Home Assistant Reverse Proxy

30 May 2026 · ~7 min read

How I think about ingress routing, automatic SSL certificates, and keeping a self-hosted platform clean without turning it into a brittle snowflake.

Read Article on Medium
Platform / Networking

Tailscale on Proxmox: Solving the “UDP GRO Forwarding” Warning

24 Jan 2026 · ~5 min read

A focused writeup on resolving host-level UDP GRO offloading bottlenecks on Proxmox VE hypervisors.

Read Article on Medium
DNS / Infrastructure

Setting Up Unbound On Raspberry Pi for Secure DNS Resolution

08 Jan 2025 · ~5 min read

DNS as core infrastructure: setting up recursive resolution, DNSSEC validation, and query privacy.

Read Article on Medium

AI & Vector Systems

AI / Multimodal RAG

Building My Own Homelab Copilot: Multimodal Chunking, Hybrid Embeddings, and Qdrant RRF Fusion

16 Sep 2026 · ~9 min read

Deep dive into engineering custom chunkers (Markdown, Excel, Draw.io XML, Loki logs), dense + sparse hybrid vector search, Qdrant RRF fusion, and cross-encoder reranking for private infrastructure.

Read Article on Medium

Automation & CI/CD

Automation

Build a Self-Updating Media Library: qBittorrent to Jellyfin Automation

19 Dec 2025 · ~6 min read

Connecting download pipeline triggers to media server updates with a focus on low-maintenance automation.

Read Article on Medium
CI/CD & Delivery

Effortless CI/CD Deployment for Homelab Applications Using GitHub Actions

14 Dec 2024 · ~6 min read

Repeatable application delivery automation with clear secrets management and reliable deployment steps.

Read Article on Medium
IaC / Documentation

Terraform Docs Tutorial: Generate READMEs for Your Modules

26 Mar 2024 · ~4 min read

Infrastructure code documentation automation: generating clean runbooks and README files directly from module inputs and outputs.

Read Article on Medium

Security & PKI

Security / Kubernetes

Secure Your Home Server: Complete Guide to Self-Signed Certificates with Kubernetes

11 Aug 2024 · ~6 min read

Treating internal PKI, trust anchors, and self-signed certificates as proper infrastructure boundaries.

Read Article on Medium
Capabilities & Credentials

Skills & Industry Certifications

Cloud Infrastructure

Azure GCP Landing Zones VNet Peering

IaC & Automation

Terraform Ansible GitHub Actions Terraform Docs

Containers & GitOps

Kubernetes Docker Docker Compose Proxmox VE

Networking & Ingress

OPNsense Traefik Tailscale Unbound DNS Outbound NAT

AI & Vector Systems

Qdrant Vector DB Hybrid Search (BM25 + Dense) RRF Fusion Cross-Encoders Ollama / Local LLMs

Observability & Logs

Grafana Loki Promtail Prometheus Frigate NVR
Certifications & Languages
Google Cloud Certified Partner Specialist: Gemini Enterprise Google Cloud Professional DevOps Engineer Google Cloud Associate Cloud Engineer Microsoft Azure Administrator Associate (AZ-104) IELTS C1 Band Goethe German A1
Contact

Let’s build reliable infrastructure & AI systems together.

Open for engineering roles, technical collaboration, or infrastructure discussions.

LinkedIn linkedin.com/in/gotoansh
GitHub github.com/AcidBurn18
Medium teckdebate.medium.com
Email Available on request / via LinkedIn