An open API service indexing awesome lists of open source software.

https://github.com/icemap/icemap

Homepage
https://github.com/icemap/icemap

Last synced: 4 months ago
JSON representation

Homepage

Awesome Lists containing this project

README

          


Hi there, I'm Qizhi Wang (Icemap)



Senior Software Engineer @ PingCAP (TiDB) | AI & Database Kernel Researcher

















### 🚀 About Me

I bridge the gap between **Distributed Systems** and **Neuro-Symbolic AI**. My goal is to mitigate hallucinations in LLMs by integrating **Knowledge Graphs** with distributed vector storage, moving from heuristic RAG to verifiable, structure-aware retrieval.

* 🔭 **Currently**: Architecting AI-native data structures at **TiDB**.
* 🏛️ **Community**: Serving as **Chair** for the *International Conference on Intelligence (ICI 2026)*.
* 🧠 **Research Interest**: Graph-Augmented RAG, Vector Search Kernels (HNSW/IVF), Distributed Consensus (Raft/Paxos).

---

### 📊 GitHub Stats


Icemap's GitHub Stats
Icemap's Top Langs

---

### 📝 Selected Publications & Manuscripts (2025-2026)

> *Focus: GraphRAG, Distributed Consensus, and Database Internals.*

* **SPRIG: Democratizing GraphRAG** - Linear, CPU-Only Graph Retrieval (Submitted to *KAIS*, 2025).
* **BALLAST: Bandit-Assisted Learning** - Latency-Aware Stable Timeouts in Raft (Submitted to *IEEE TPDS*, 2025).
* **TiCard: Deployable EXPLAIN-only Residual Learning** - Cardinality Estimation (Submitted to *JDSA*, 2025).
* **RIFT: Replicating Ideological Fractures via Transformers** (Submitted to *JASSS*, 2025).

---

### 🛠️ Engineering Highlights

#### **PingCAP (TiDB)** | *Senior Software Engineer (AI & Ecosystem)*
* **Hybrid Query Optimization**: Formulated cost models for Hybrid Search (Vector + SQL) within the TiDB kernel.
* **Ecosystem**: Ported Hibernate ORM to TiDB and aligned client SDKs with the Percolator transaction model.

#### **Tencent Music Entertainment (TME)** | *Former Senior Engineer*
* **High Concurrency**: Designed the *TME Live* platform handling **10M peak QPS** using Bloom Filters/HyperLogLog for massive dataset distinct counting.

---

### 💻 Tech Stack

| Domain | Technologies |
| :--- | :--- |
| **Languages** | ![Go](https://img.shields.io/badge/-Golang-00ADD8?logo=go&logoColor=white) ![Java](https://img.shields.io/badge/-Java-007396?logo=java&logoColor=white) ![Python](https://img.shields.io/badge/-Python-00599C?logo=python&logoColor=white) |
| **Systems** | ![TiDB](https://img.shields.io/badge/-TiDB-444?logo=tidb) ![MySQL](https://img.shields.io/badge/-MySQL-4479A1?logo=mysql&logoColor=white) ![Raft](https://img.shields.io/badge/-Raft_Consensus-E535AB) ![Redis](https://img.shields.io/badge/-Redis-DC382D?logo=redis&logoColor=white) |
| **AI & Data** | ![HNSW](https://img.shields.io/badge/-HNSW_Indexing-success) ![GraphRAG](https://img.shields.io/badge/-GraphRAG-blueviolet) ![Vector DB](https://img.shields.io/badge/-Vector_Database-ff69b4) |




Let's connect! I'm open to discussing Neuro-Symbolic AI, Database Kernels, or just geek out on Golang.