Systems & AI Infra Engineer. Building high-performance storage systems and studying distributed training frameworks
📍 Wuhan · ✉️ guozhihaoemail@gmail.com
| Project | Description |
|---|---|
| AReaL | Distributed RL training framework for LLM alignment — async rollout, FSDP2/Megatron backends, SGLang/vLLM inference |
| kiwi-rs | Redis-compatible elastic KV storage built on RocksDB with Raft consensus — Rust rewrite of kiwi-cpp |
| kiwi-cpp | The original C++20 Redis server implementation on RocksDB |
| rrproxy | Lightweight Rust Redis proxy with multi-tenant and multi-library support |
- Distributed training systems — parallelism strategies, GPU cluster scheduling, RL at scale
- High-performance storage — LSM-tree engines, Redis internals, proxy & sharding
- Systems programming — lock-free data structures, async runtime, memory management