Projects

Built this to understand continuous batching properly. Supports priority-based scheduling, prefix caching, and has a terminal dashboard for monitoring request queues and GPU utilization in real time. Most inference servers are either closed-source or designed for Google-scale — this one is meant to be readable.

Interactive learning platform for system design. Includes a working Kafka simulator and D3-based concept graphs. Built it because most system design resources describe things without ever showing them — used it myself for interview prep.

Terminal interface in Next.js. Supports filesystem navigation, inline blog rendering, and a 3D force-directed graph of GitHub activity. Made it this way because the standard portfolio template is deeply boring to build and probably to read.