Builder, operator2024–PresentSmart home + AI cameras
Home Assistant Frigate Docker Object Detection
Home Assistant
Home Assistant runs on the homelab and controls everything that plugs in: lights, switches, sensors, and automations.
It replaces the proprietary apps that come with each device. Instead of five different apps for five different brands, there’s one dashboard. Automations work across brands — motion sensor from one brand can trigger lights from another.
The setup is straightforward: Zigbee USB dongle for most devices, a few Wi-Fi ones that don’t have alternatives, and Home Assistant handles the protocol translation.
Frigate
Frigate is the camera system. It does object detection directly on the camera feeds — people, dogs, cars — and only alerts when something relevant happens.
The detection runs locally, not in the cloud. That means:
No monthly subscription for smart detection
No video footage leaving the network
Alerts are instant — there’s no round-trip to a server
It uses hardware acceleration where possible and falls back to CPU otherwise. The detection is good enough that false positives are rare — it distinguishes between people, animals, and vehicles without much tuning.
How They Work Together
Home Assistant listens to Frigate events and triggers automations:
Person detected at the front door → lights turn on, camera feed shows on the TV
Car detected in the driveway → notify that we’re home
No motion for X hours → cameras switch to low-frame-rate mode
The camera system feeds into the homelab’s observability stack too — uptime, storage usage, and detection accuracy all show up in Grafana.
What’s Running
Several IP cameras around the house
Zigbee sensors (motion, door, temperature)
Smart lights and switches
Frigate with object detection models
Home Assistant as the orchestrator
It’s not the most sophisticated setup, but it’s the kind of thing that makes you forget about the infrastructure until something goes wrong. And when it does, at least I can fix it.
{"menu":[{"name":"Pages","items":[{"label":"Home","subtitle":"Overview","action":"navigate:/","icon":"page"},{"label":"Career","subtitle":"Timeline & principles","action":"navigate:/career","icon":"page"},{"label":"Projects","subtitle":"All projects","action":"navigate:/projects","icon":"page"},{"label":"About","subtitle":"About Meher","action":"navigate:/about","icon":"page"}]},{"name":"Settings","items":[{"label":"Toggle Theme","subtitle":"","action":"toggleTheme","icon":"theme"}]}],"fuse":{"threshold":0.6,"minMatchCharLength":2,"keys":["label","subtitle","searchableText"]},"projects":[{"title":"Global Payroll Platform","description":"In-house payroll platform across 6 APAC markets, $1.2B annually. I define the technical requirements, coordinate payments integrations, and make sure the engineering teams build what the business actually needs.","tags":["Python","SQL","API Integration","ISO 20022"],"link":"#","image":null,"techStack":["Python","SQL","API Integration","ISO 20022"],"size":"large","domain":"fintech","icon":null,"featured":true},{"title":"Fraud Detection Engine","description":"1.5M payment transactions analyzed. Built automated detection rules that caught the obvious cases — the ones that shouldn't need a human reviewing them. Cut manual validation by 80%. QuickSight dashboards so the non-technical teams could see what was flagged and why.","tags":["Python","QuickSight","Analytics"],"link":"#","image":null,"techStack":["Python","QuickSight","SQL"],"size":"medium","domain":"analytics","icon":null},{"title":"Background Check Revamp","description":"India's background checks were taking four months at the 90th percentile. Candidates would leave. Worked with Legal, Compliance, and Business to renegotiate vendor SLAs, parallelize the workflow, and grow the team from 10 to 45. Cut it to one month, impacted 120K annual hires.","tags":["Operations","Compliance","Process Engineering"],"link":"#","image":null,"techStack":["SQL","QuickSight","VBA"],"size":"medium","domain":"data","icon":null},{"title":"Sovereign Homelab","description":"50 services on bare metal: reverse proxy, DNS, VPN, password manager, photo library, media streaming, AI inference, git CI/CD, smart home, Matrix federation. Full observability via Grafana, Loki, Prometheus, and Alloy. No cloud provider.","tags":["Docker","Linux","Caddy","NetBird"],"link":"#","image":null,"techStack":["Docker","Caddy","NetBird","Vaultwarden"],"size":"large","domain":"infrastructure","icon":null,"featured":true},{"title":"AI Node","description":"Private LLM inference on CUDA GPU. llama.cpp server with quantized models, Open WebUI frontend. Used through OpenCode on my laptop and my Hermes agent on Matrix when I'm out. No cloud APIs, no telemetry.","tags":["CUDA","llama.cpp","Open WebUI"],"link":"#","image":null,"techStack":["Python","CUDA","llama.cpp"],"size":"medium","domain":"ai","icon":null,"featured":true},{"title":"AI & LLM on Your Homelab — A Tutorial","description":"Step-by-step guide to running your own AI inference stack at home. llama.cpp, Open WebUI, Caddy reverse proxy, Matrix integration — everything local, everything yours.","tags":["Tutorial","AI","llama.cpp","Homelab"],"link":"#","image":null,"techStack":["llama.cpp","Open WebUI","Caddy","Docker"],"size":"medium","domain":"ai","icon":null,"featured":false},{"title":"Matrix Citadel","description":"Self-hosted Matrix federation with full voice and video via dual LiveKit servers. Twunnel bridges between two homeservers, JWT auth services, OpenClaw Gateway for AI agent orchestration.","tags":["Matrix","LiveKit","Twunnel"],"link":"https://cloudcitadel.in","image":null,"techStack":["Matrix","LiveKit","Twunnel","OpenClaw"],"size":"medium","domain":"infrastructure","icon":null},{"title":"Media & Photo Stack","description":"Jellyfin streaming with automated library management, subtitles, and analytics. Immich photo library with ML-powered face recognition and reverse image search.","tags":["Jellyfin","Immich","Media Server"],"link":"#","image":null,"techStack":["Jellyfin","Immich","PostgreSQL"],"size":"medium","domain":"infrastructure","icon":null},{"title":"Observability Stack","description":"Full infrastructure telemetry: Grafana dashboards, Loki log aggregation, Prometheus metrics, Alloy collector, plus node, process, and disk health exporters. Monitoring 50 services and bare-metal health in real time.","tags":["Grafana","Prometheus","Loki","Alloy"],"link":"#","image":null,"techStack":["Grafana","Prometheus","Loki","Alloy"],"size":"medium","domain":"infrastructure","icon":null},{"title":"Gitea","description":"Self-hosted git server with CI runner and built-in container registry. This is where the folio lives — code pushed here, the runner builds it, the registry stores the image, and it deploys back to the same machine. Source to production, nothing leaves the box.","tags":["Git","CI/CD","Container Registry"],"link":"https://git.nexusno.de","image":null,"techStack":["Gitea","Docker","CI/CD"],"size":"medium","domain":"infrastructure","icon":null},{"title":"Immich","description":"Photo library with ML-powered face recognition, reverse image search, and automatic duplicate detection. Replaces Google Photos without the telemetry — all the smart features, none of the cloud.","tags":["Immich","ML","Photos"],"link":"#","image":null,"techStack":["Immich","OpenVINO","PostgreSQL"],"size":"medium","domain":"infrastructure","icon":null},{"title":"Home Assistant + Frigate","description":"Smart home automation with AI security camera system. Frigate does hardware-accelerated object detection — people, dogs, cars — on the cameras, not the cloud. Home Assistant ties everything together: lights, sensors, automations, and camera events.","tags":["Home Assistant","Frigate","Smart Home"],"link":"#","image":null,"techStack":["Home Assistant","Frigate","Docker"],"size":"medium","domain":"infrastructure","icon":null},{"title":"SearXNG","description":"Self-hosted metasearch that aggregates results from Bing, Wikipedia, and others without tracking. Replaces the browser search bar — same results, no data broker on the other end.","tags":["SearXNG","Privacy","Search"],"link":"https://search.nexusno.de","image":null,"techStack":["SearXNG","Docker","Valkey"],"size":"small","domain":"infrastructure","icon":null}],"skills":{"tools":["Python","SQL","QuickSight","VBA","Docker","Linux","Kubernetes","Grafana","Prometheus","Loki","Caddy","Matrix"],"standards":["ISO 20022"],"domains":["Payments","Infrastructure","Process Engineering"]},"contact":{"email":"hi@meherchaitanya.com","channels":[{"label":"Email","url":"mailto:hi@meherchaitanya.com","displayText":"hi@meherchaitanya.com","icon":"mail","external":false},{"label":"LinkedIn","url":"https://linkedin.com/in/meherchaitanya","displayText":"meherchaitanya","icon":"linkedin","external":true},{"label":"Matrix","url":"https://matrix.to/#/@meher:hanumara.online","displayText":"@meher:hanumara.online","icon":"matrix","external":true}]}}