Building a Weighted Carpool Ranking Engine
Finding nearby carpools is only half the problem. Delivering the right one first requires combining multiple business signals into a single relevance score.
Notes on distributed systems, API design, and the engineering decisions behind production systems.
Finding nearby carpools is only half the problem. Delivering the right one first requires combining multiple business signals into a single relevance score.
How Redis geospatial indexing enabled fast nearby driver discovery without relying on expensive database geospatial queries.
Knowing who is online sounds simple until every feature depends on it. This article explores how Redis-backed presence powered ride dispatch, messaging, availability, and real-time user experiences.
Marketplace-style ride requests introduce a different concurrency problem. This article explores how ride assignment was designed to guarantee that only one driver's offer can ever become the final booking.
When multiple passengers attempt to book the last available seat simultaneously, correctness matters more than speed. This article explores the concurrency strategy behind preventing overbooking.
Real-time applications cannot assume a socket connection will remain alive. This article explores how session recovery, room reconstruction, and state synchronization allowed drivers and passengers to seamlessly continue active rides after reconnecting.