A horizontally scalable WebSocket sync engine — presence, live updates, and offline replay for collaborative apps.
Collaborative apps quietly assume realtime — live cursors, shared state, instant notifications. But rolling your own WebSocket layer usually ends in flaky connections, missed messages, and no story for reconnects.
A naive socket layer has three failure modes: reconnects lose the messages missed while offline, one chatty room can stall the whole process, and scaling out means duplicate or dropped fan-out.
Designed a connection gateway that keeps state out of the socket — presence and rooms live in Redis.
Routed events over Redis pub/sub so any node can serve any client and fan-out stays consistent.
Built an outbox and replay so a client that drops misses nothing; state reconciles on reconnect.
Added backpressure and per-room limits so a noisy room can't stall the process.
The socket is just a door. The system lives in Redis — presence, rooms, and an outbox that guarantee a reconnecting client catches up to the exact message it missed.
10k+concurrent clients
on a single node — fan-out over Redis pub/sub, replay on reconnect
RAG-grounded LLM workflows that answer enterprise questions from real company knowledge — not a demo corpus.
100% grounded — every answer traced to a retrieved source — never generated from memory
The REST APIs and workflows behind Peoplebox's core HR modules — one service layer powering six product areas.
6 product areas — Goals, Reviews, Surveys, HRIS, IDPs, and reporting on one service layer
EXPLAIN-first schema and query work that kept enterprise dashboards and workflows fast as data grew.
p95 response time — flattened on million-row tables by index-backed, EXPLAIN-first query plans