About
I'm Mirtunjay. I build data systems.
By day I'm an AVP–Data Engineer at an MNC — Spark, Databricks, pipelines, migrations, and the 3 AM pages that teach you what "done" really means. By night I build systems of my own: a study platform, a wealth tracker, a mutual-fund intelligence system, and the tooling around them. This notebook is where that work is recorded — not as a highlight reel, but as evidence.
Every claim here traces back to something I can point at: code, tests, logs, measurements. Where something isn't measured, the page says so. Where something broke, the page keeps the failure on record next to the fix. I hold outside work to the standard I'd defend in any review, and this site is where I prove it — project by project, incident by incident.
Lately the work changed shape. AI writes most first drafts now, including some of the code on this site. So the notebook changed with it: less "here is code," more "here is the reasoning the code came from" — constraints, rejected alternatives, verification, and the failures kept on record. Implementation is cheap. Judgment is the scarce part, and this is where I practice it in public.
What I work on
- Data engineering — Spark, Databricks, pipelines, migrations
- AI and agents — LLM systems, memory, evaluation, tooling
- Systems and architecture — storage, sync, offline-first design
- Developer tools — scanners, notebooks, automation that compounds
How I work, in one page
Reason first, then build, then measure. Decisions get written down with their alternatives. Tests pin the math. Metrics come from runs, not adjectives — and the gaps are labeled, not hidden. If you want the method stated plainly, open any project page and read its struggle section first: that's the notebook in miniature.