Hi, I'm
Auxten Wang
AI & Database Guy
Technical Director at ClickHouse. Creator of chDB — an in-process OLAP engine that was acquired by ClickHouse in 2024. I build databases, recommendation systems, and tools that make complex things simple.
Hi, I'm
AI & Database Guy
Technical Director at ClickHouse. Creator of chDB — an in-process OLAP engine that was acquired by ClickHouse in 2024. I build databases, recommendation systems, and tools that make complex things simple.
In-process OLAP SQL engine powered by ClickHouse. Supports 60+ data formats, 334 SQL functions, and runs embedded in Python, Go, Rust, Node.js, Bun, and more — zero infrastructure needed.
A decentralized, Byzantine Fault Tolerant SQL database with blockchain features. DPoS + BFT-Raft consensus, immutable audit trail, column-level ACL — an open source alternative to Amazon QLDB.
A pure Go SQL database in ~2000 lines. B+Tree storage, volcano model planner, SQL parser — built from scratch for database theory research.
Pure Go PostgreSQL parser supporting SQL:2011 down to SQL-92. Used by Atlas, ByteBase, and other production tools.
Deep learning recommendation framework in Go. DeepFM, DNN, Deep Interest Network — with Item2vec embeddings and MovieLens benchmarks.
Slack-like messaging for AI agents with E2E encryption. Channels, sessions, threads, cron scheduling, debug panel — on Web, macOS, and iOS.
Menu-bar companion for headless Mac deployment. Virtual display, audio router, awake manager, system monitor — built for AI agent operations.
Three-layer memory system for AI agents. Hybrid semantic + keyword search, local-first with LAN sharing, built on chDB and local embeddings.
Interactive 3D tour of how LLMs run on real hardware. Three zoom levels — datacenter, NVL72 rack, chip — for training, inference, MoE, KV cache, and failure recovery.
I'm Auxten Wang, a database engineer and open source builder based in Singapore.
I created chDB — an embedded ClickHouse engine that lets you run full OLAP SQL anywhere without infrastructure. It started from a simple need: running ClickHouse in Jupyter Notebooks for ML model training instead of waiting on slow Hive clusters. chDB was acquired by ClickHouse Inc. in 2024, and I now work there full-time growing the project and its ecosystem across 7+ language bindings.
Before ClickHouse, I was Principal Engineer at Shopee, Cofounder & CTO of CovenantSQL (decentralized BFT database, 1,500+ stars), and RecSys & Infra Director at 4Paradigm. Earlier in my career, I was Technical Director at eLong, built cluster control for 30,000 servers at Qihoo 360, and created Gingko — Baidu's P2P dataset distribution system — writing 90% of the C++ code. Started my journey as an intern at Douban in 2009.
Lately, I've been exploring the intersection of databases and AI agents — building memory systems, RAG pipelines, and MCP integrations powered by ClickHouse.
chDB acquired. Building the embedded ClickHouse ecosystem full-time, remote from Singapore.
ML platforms, databases, and recommendation systems at scale. Beijing.
Built a decentralized BFT SQL database — open source Amazon QLDB alternative. 1,500+ GitHub stars.
Led recommendation systems and infrastructure at one of China's leading AI companies.
Led engineering teams. Founded eLong's internal tech university.
Ops dev, wireless security research. Created 360 Traffic Guard. Designed cluster control for 30,000 servers.
Built Gingko — Baidu's P2P large dataset distribution system in C++ with libevent. Designed architecture and wrote 90% of the code.
Built an automated web testing framework with Python, Selenium, and Xvfb.