How to design ledger systems that survive concurrency, partial failures, and audits. Covers double-entry invariants, event-sourced architectures, idempotency, and reconciliation patterns used in production.

Ship a real webhook delivery service with idempotency, retries, and a dead-letter queue — a portfolio project that signals genuine distributed-systems skill to hiring managers.

Compare io_uring and epoll as Linux async I/O architectures, covering kernel internals, syscall costs, and the production patterns where each one wins.

How BBR actually estimates bottleneck bandwidth and RTT, why v2/v3 fix v1’s loss-insensitivity, and what to tune at the kernel, application, and NIC layers before you ship it.

A working engineer’s guide to QUIC stream multiplexing: how independent byte streams coexist over a single connection, why per-stream flow control fixes TCP’s head-of-line blocking, and where it still bites.