A Rust-based streaming image processor that compares naive, batched, and pipelined approaches to high‑volume image resizing. It uses bounded channels and semaphores to control backpressure and in‑flight work, achieving major throughput gains while tracking peak memory to avoid OOM scenarios.

The pipeline processes images from Lorem Picsum and reports per‑stage timing, throughput, and memory stats, with structured logging, benchmarking, and test coverage. I also wrote a blog post documenting the motivation, architecture, and results here.