{
"name": "Bing Wang",
"alias": "Wayne",
"location": "Berlin, Germany (Open to relocation)",
"status": "Available immediately",
"experience": "15+ Years (10 years at Alibaba & Ant Group)",
"domains": [
"High-Availability Platforms (99.999%)",
"AI & RAG Pipelines over Unstructured Data",
"Game Engine & Open-Source Development"
]
}| Layer | Technologies |
|---|---|
| Frontend | |
| Backend | |
| AI & Data | LiteLLM RAG Pipelines |
| Infra & Cloud |  |
📚 DictClub
A smart vocabulary learning platform that bridges reading and acquisition.
- 🏗 Architecture: React (SSR & SEO Optimization) + MariaDB + Multi-source Data Cleaning Pipeline + Chrome Extension Worker
- 💡 Highlight: Indexed 200k+ dictionary entries with contextual word-sense resolution. Features spaced-repetition scheduling and adaptive item selection.
High-performance MMO city-building simulation engine pushing the boundaries of deterministic sync and zero-GC optimization.
- 🏗 Architecture: Highly decoupled Skynet (Server) and SDL3/C++17 (Client) architecture, driven by LuaJIT. Utilizes KCP (Reliable UDP) + Sproto for ultra-low latency, paired with a robust client Jitter Buffer.
- ⚡ Synchronization: Hybrid Lockstep & State-Sync model via a shared Server-Authoritative ECS. Enforces strict mathematical determinism across identical Lua ECS pure functions. Features 60-frame state checksums with seamless
dirty_entitiesdelta resync and Optimistic UI prediction. - 🚀 Performance: Zero-GC optimization at the architectural level. Employs in-place table modifications across the C++/Lua sol2 boundary and Lua FFI C-level pointer offsets for network decoding. Algorithmic bottlenecks strictly constrained (e.g., O(logN) QuadTrees), all while supporting instant Lua hot-reloading.
