I’m Sandeep — a senior software engineer at LinkedIn working on distributed systems and large-scale infrastructure. I’ve spent the last several years shipping the kinds of services that fail in interesting ways at 3 AM, and the notes here are my attempt to keep the lessons honest.
Why this site exists
Most engineering writing online is either tutorial-shaped (here’s how to configure X) or résumé-shaped (here’s how my team scaled Y). Both are useful, but neither is the kind of thing I personally re-read. What I keep coming back to are notes built from first principles — where the why comes before the how, where the trade-off is named before the solution, and where the historical context (the original paper, the bug that motivated the design) is given the space it deserves.
This site is a curated slice of my private notebook. Many drafts and half-formed ideas stay unpublished by design; I’d rather publish twenty notes I’m willing to defend than two hundred I’m not.
What I write about
- Distributed systems and databases — the primitives, the failure modes, and the trade-offs that only become obvious in production. Consensus, MVCC, CDC, consistent hashing, the usual cast.
- System and architecture design — event sourcing, BFF, multi-tenancy, strangler-fig migrations. Patterns named for the problem they solve, not the framework that popularised them.
- AI systems in production — RAG, model serving, embeddings, vector stores. Pulled from shipping ML in real environments, not from the model card.
How this site is built
It’s a Quartz site rendered from an Obsidian vault — Obsidian is the source of truth, every diagram is an Excalidraw drawing rendered to dark/light SVG pairs, and the entire build runs from a small set of scripts in the public repo. If you spot a bug or a stale link, the PR queue is open.
Get in touch
If a published note is wrong, unclear, or contradicts your production experience, I’d genuinely like to hear about it. The fastest path is to open an issue on GitHub.