I'm a freelance DevOps & Platform Engineer. I take platforms from an empty cloud account to production and then keep them alive — infrastructure, pipelines, the backend services that run on top, and the observability that tells you when any of it is lying to you.
I work across several orgs at once, so a given day might be a Terraform module, a Kubernetes upgrade, a FastAPI service, or a 3 a.m. alert that turned out to be DNS. Right now I'm going deeper on LLM infrastructure and MLOps — vector stores, embedding pipelines, and getting AI systems to behave in production.
AWS Certified Cloud Practitioner · Microsoft SC-900 · Microsoft AZ-900
Building — Smriti · स्मृति
A fully offline photo library for the folders already sitting on your machine — the parts of Google Photos you actually miss, with nothing ever leaving your laptop.
- On-device face recognition — SCRFD + ArcFace, CPU only
- Offline reverse geocoding, so Places works with the network unplugged
- Auto-detected trips, duplicate cleanup, Google Takeout repair
- FastAPI + React, shipped as a Tauri desktop app for macOS & Windows
brew install jadonharsh109/tap/smritiAlso building — Advect
A plain-English prompt in, an animated GIF of a request flow out — infrastructure diagrams that move, generated and rendered with no cloud API key anywhere in the loop.
- Runs on a local model through Ollama — nothing leaves the machine, no per-render cost
- BM25 retrieval hands the model the nearest hand-tuned pattern to adapt, not author blind
- Validators and a critic feed a repair loop — structurally valid but unreadable output gets rejected
- JSON IR → lane layout → animated SVG → Playwright capture → ffmpeg, byte-identical under golden tests
advect prompt "explain how a canary release shifts traffic to a new version"Most of my work happens in client and organisation repos rather than here.



