Chao

I have spent more than 20 years working across software engineering, large-scale systems, and technical leadership — while staying hands-on with system design, engineering experiments, and code.

My experience spans embedded systems, high-concurrency internet services, large enterprise systems, and more recently LLMs and agent-based workflows.

What I tend to work on

Most of my work sits somewhere between engineering, architecture, product, execution and organization.

I am especially interested in problems that are difficult to understand from only one of those perspectives.

Complex systems and architecture

Systems that have accumulated multiple generations of architecture, business constraints and operational history, where the right next step is not simply a rewrite or another layer of abstraction.

Complex projects and cross-functional execution

Projects involving multiple systems, teams or business functions, where progress depends as much on problem definition, interfaces and responsibility boundaries as on implementation.

AI and agent workflows

I have been actively integrating LLMs, coding agents, multi-agent systems and automation into real engineering and research workflows.

The focus is less on isolated demos and more on making these systems useful over long-running, complex tasks.

Technology and business decisions

Technical feasibility is only one part of a decision.

Cost, alternatives, organizational effort, business logic, risk and long-term return often matter just as much.

What I am working on now

I maintain my own technical lab and continue to build systems that I use in real work.

AI systems

Running and evaluating local and cloud models across multi-GPU NVIDIA systems, Apple Silicon and AMD hardware.

Current work includes inference frameworks, quantization, Dense and MoE models, long context, KV cache, TTFT, prefill performance, concurrency and deployment trade-offs.

Agent systems

Building agent workflows for software development and long-running technical tasks.

This includes main/sub-agent coordination, sandboxing and permissions, context compression, handoff, memory, RAG, automated verification and failure recovery.

Research systems

Building long-running systems for company and financial research, combining documents, financial data, event streams and AI-assisted analysis.

The goal is not just information retrieval, but maintaining hypotheses, comparing companies, checking facts and supporting decisions over time.

Background

Over the years I have worked with systems ranging from embedded software and high-concurrency internet platforms to games, video, e-commerce, logistics and supply-chain systems, transportation systems, retail platforms, procurement systems and ERP.

As systems and organizations grew, my role increasingly crossed engineering, architecture, product, project execution and organizational design.

I have also built two multidisciplinary teams from scratch to more than 50 people.

My academic background includes graduate degrees in Software Engineering and Finance from two of China's C9 universities, as well as CFA Level I and Level II.

Finance has never been a second career. It mainly shaped the way I think about opportunity cost, alternatives, risk, reversibility, business models and long-term value.

More recently, LLMs and agents have become another layer in the same progression: new tools to extend how engineering, research and complex knowledge work can be done.

Open to interesting conversations

I am open to projects, advisory work, part-time or remote collaboration, longer-term partnerships, and selected full-time opportunities where there is a strong mutual fit.

The common thread matters more than the formal arrangement: an interesting problem, enough complexity to be worth understanding properly, and people who can work through it directly.

This site intentionally does not include my complete employment history, company names or identifiable project details. I am happy to share relevant background when there is a real reason to continue the conversation.

Contact: chao@geeke.tech