Systems & AI Infra Engineer · Building high-performance storage systems and distributed training frameworks
|
Distributed RL training framework for LLM alignment — async rollout, FSDP2/Megatron backends, SGLang/vLLM inference.
|
Redis-compatible elastic KV storage on RocksDB with Raft consensus — Rust rewrite of kiwi-cpp.
|
|
C++20 Redis server implementation on RocksDB. |
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