system.status: online
Hou Huiyang
侯惠阳Long-term practice across technology, product, and organizations
Engineer by training, product thinker by habit, organizational builder by experience. I care about getting AI into real workflows—making organizations more effective without losing human judgment or creativity.
Methods & Systems
Frameworks in Practice
AI-Native Capability Graph
Not AI as a helper — AI as the execution unit
AI Coding is the execution layer, AI Team Enablement makes individuals more effective, and AI Team Collaboration keeps the organization in sync. The three reinforce each other. Select any node for details.How I Think
Ideas I Keep Returning To
Systems Thinking
Find structure and feedback before reaching for tools. Complex problems rarely lack methods—they lack leverage.
In AI-native work, map capability layers and business loops first, instead of stacking models and assistants.
Engineering Reality
Abstractions must become reusable standards. A demo is not done until others can build on the capability.
After a scenario validates, turn it into platform capability and workflow—not a one-off pilot.
Human + AI
AI extends execution; people set goals, boundaries, and judgment. The aim is amplification, not replacement.
In multi-agent systems, human takeover is part of the quality loop—not an afterthought.
The Road So Far
Journey Highlights
CTO
Chengdu Mengxiang Technology (Shixiseng)Leading an AI-native transformation and building a shared capability foundation across technology, product, and organization.
Founder / CTO / Director of Engineering
Xinxu Cloud Brain AI · ZhonglebangBuilt intelligent products and payment systems from zero to one, sharpening judgment through startups and scaled delivery.
Technical Expert / Product & Engineering Lead
DiDi · Baitong Century · DataokeBuilt engineering systems for complex, high-concurrency businesses and helped key products evolve reliably.
Algorithm Strategy Engineer / Technical Lead
Baidu · Xiaomi · Meituan · Oak PacificDeveloped engineering and architecture foundations across diverse teams and core production systems.
AI Practice
Hands-on, not slides
Orchestrating model APIs into shippable agents
AI Agent Engineering
Not just calling models—composing tool use, RAG, structured outputs, and workflows into agents that connect to real business, can be evaluated, and keep improving.
From random weights to local inference
Training a Transformer from Scratch
Implementing the full pipeline: dataset generation, tokenization, a decoder-only Transformer, pretraining, instruction tuning, evaluation, checkpointing, export, and local inference.
Engineering Notes
Long-form thinking in practice
Turning practice into ideas that can be reused, challenged, and improved.
View all notesAgent Architecture for SMBs: A 90-Day Path to Production
Do not copy an enterprise AI platform. Build a production agent with deterministic orchestration, narrow tools, least privilege, evaluation, and measurable business outcomes.
Building Docket with Pragmatic DDD
A code-based look at how Docket uses bounded contexts, domain services, and a modular monolith to manage a professional-services workflow—and how the architecture can evolve.