vik

teaching databases to do things nobody asked for.


i'm 19, from india, building toward database infrastructure work, mostly distributed systems and the durable, low level stuff that decides whether a system actually survives a crash.

right now that's quorum loro gateway, a crdt sync server built on top of a quorum replicated durable log, postgres cdc, a connector moving committed wal into feldera over http, and seam, backfilling old postgres rows into a second database without missing the live writes that land mid copy.

taught myself most of this by picking a real system, poking at its internals until something broke, then figuring out why.

writeups

01real-time data replication & its caveatswhat hides behind the arrow between a source and a destination: acks, crash windows, retention cliffs, and backfill seams.postgres · kafka · cdcseptember 202602the cost of saying safemoving committed postgresql changes into feldera over http, acked only after a completion token.rust · postgres · felderaseptember 202603quorum loro gatewaya durable crdt sync gateway for loro, acknowledging writes only after quorum commit on ursula.rust · ursula · loro protocoljuly 2026