Developed a fault-tolerant, distributed consensus system from scratch using the Raft algorithm, ensuring strong consistency and high availability by automatically electing new leaders in under 80ms for recovery from node crashes, guaranteeing cluster integrity and zero data loss. Implemented core mechanisms like leader election and log replication using gRPC, and integrated the Pebble KV store with Write-Ahead Logging (WAL) for durable state persistence.
Engineered a horizontally scalable, distributed key-value store from the ground up, implementing replication and sharding strategies to ensure high availability and fault tolerance. Implemented a self-healing system with a gRPC control plane that automates leader failover and state recovery in under 4 seconds, using WAL and Quorum protocols to ensure strict data consistency.
Developed a CPU scheduler with five distinct policies and engineered core concurrency primitives including semaphores, shared memory (IPC), and a user-level thread library. Implemented an MLFQ scheduler, reducing average response time for I/O-bound processes by 73% over FIFO and enabling complex multi-threaded applications to run efficiently.