Hi, I’m cmkhoa.

A Redis-compatible server in Go with RESP2/TCP, a 32-shard concurrent keyspace, expiration and eviction policies, AOF persistence, and race-tested multi-key operations.

A local Go service for supervising llama.cpp inference with model registration, request queueing, streaming, health checks, restart recovery, and operational metrics.

A reproducible research pipeline for LLM-assisted grading of worked calculus solutions, with versioned prompts, family-disjoint splits, schema validation, and structured metrics.