I build infrastructure for AI agents and quantitative systems.

My work is centered on one question:

How do we make autonomous systems reliable enough to operate in complex, high-stakes engineering environments?

I focus on modular system design, evaluation, monitoring, replay testing, and operational safety. My current projects involve AI agent workflows, quant research infrastructure, market data pipelines, and execution systems.

• Agent workflow design
• Tool-use architecture
• Evaluation and scoring frameworks
• Instruction-following and compliance measurement
• Human-in-the-loop control
• Observability, auditability, and replay

• Backtesting architecture
• Market data pipelines
• Orderbook analysis
• Signal generation and execution diagnostics
• Risk controls and system monitoring
• PostgreSQL-based research data models

• Linux / WSL / remote development environments
• Modular service design
• CI, testing, logging, metrics, rollback, and reproducibility
• Systems programming, numerical computing, and research tooling

• Rust
• C / C++
• OCaml
• SystemVerilog
• Python
• Bash
• R
• C# / .NET
• q / kdb+
• Go
• TypeScript

• Chinese
• English
• German

I study Condensed Matter Physics and Algebraic, Differential Topologuy at UC Berkeley.

I am also a classical musician and have professional experience in the German legal field.

• I love driving my old Jeep into untouched wilderness—deserts, forests, and coastlines—with my dog, a vintage large-format camera, and my diving gear (without my ThinkPad laptop)

• I am currently building a hybrid wheeled quadruped robot, which I hope will be able to carry 100 kg of diving equipment for me across rocky reefs. I am also modifying my Choptima CCR and developing a high-performance diver propulsion vehicle.

Please reach me through GitHub.