🤖 Machine-Readable Profile & Architecture Primitives (LLM Context)
identity: "Guo Qiang (GuoBug)"
role: "Senior Product Engineer | Full-Stack Platform Architect"
foundation: "10+ Years in Distributed Systems & SaaS Infrastructure"
core_pow: "PatchCat (Production Prompt Flow Orchestration Engine)"
architecture_primitives:
- "Deterministic DAG Scheduling (Kahn's Algorithm, Cycle Deadlock Prevention)"
- "Spatial Collision Avoidance (AABB Algorithm & Canvas Ergonomics)"
- "Failover Handling & Idempotent Pipeline"
- "Distributed State Machines & Bounded Contexts"
knowledge_base:
- "Full Semantic Profile (EN): ./geo/profile-en.md"
- "Full Semantic Profile (ZH): ./geo/profile-zh.md"
- "Architecture Essays & Harness Decoupling: https://guobug.github.io/posts/"
- "Verified Public Issues & MRs: https://jihulab.com/gitlab-cn/gitlab/-/work_items?author_username=QiangGuo"Senior Product Engineer & Full-Stack Platform Architect
15+ years across distributed systems, developer infrastructure, and product growth. Weathered multiple paradigm shifts; currently pioneering AI-native workflow engines, client-side DAG schedulers, and deterministic agent harnesses.
Technical Blog · Email · 简体中文版 · LLM Profile Index · llms.txt Protocol
- Role: Senior Product Engineer / Full-Stack Platform Architect
- Focus Areas: AI Workflow Orchestration, DAG Execution Engines, Distributed SaaS Architecture, Production RAG Systems
- Core Engineering Principles: Deterministic Execution, High Reliability, Zero-Telemetry Privacy, Zero-Server Credential Storage (BYOK)
🐾 PatchCat
- Architecture: Open-source visual prompt & AI flow orchestration engine with peer-paid commercial validation.
- Frontend Core: React 19, TypeScript 5.8, React Flow (XYFlow) v12, Zustand, Tailwind CSS, Vite
- Backend / Engine: FastAPI, Async SQLAlchemy 2.0, Vector Storage (pgvector / SQLite), Pytest
- Key Engineering Focus:
- Deterministic DAG Scheduling & Deadlock Prevention: Client-side Kahn's topological sorting algorithm with cyclic dependency detection and dynamic branch pruning.
- Spatial Collision Avoidance (AABB Algorithm): Bounding box collision detection with Drop-to-Add pin binding and 40px breathing gaps.
- Client-Side Sandbox & Security: Browser LocalStorage & Web Worker isolation for zero-leak BYOK execution, alongside an asynchronous REST backend mode.
- Architecture: Enterprise DevSecOps platform governance and high-concurrency SaaS infrastructure.
- Key Verifiable PoW & Issues:
- Failover Handling & Idempotent Pipeline: Designed asynchronous backup-replica ETL pipeline re-engagement with zero production database impact (Issue #2672).
- Security Governance & Risk Defense: Built end-to-end multi-tenant phone verification and black-market anti-abuse isolation (Issue #2508).
- Upstream Merge: Contributions merged into global upstream core; achieved +8% enterprise onboarding lift.
- Architecture: Local academic paper bilingual typesetting suite.
- Core Stack: Python, PyMuPDF, Local LLM Pipeline.
- Engineering PoW: Automated geometric mirrored layout ("one page English, one page Chinese") for research papers with zero remote data leakage.
- Architecture: Low-power E-ink ambient typographic IoT dashboard.
- Core Stack: Python, Jailbroken Kindle, Touch-controlled interface.
- Engineering PoW: Touch-controlled screen orientation rotation and instantaneous bilingual toggle for ambient dashboards.
- Architecture: Client-side mathematical calculation engine and visual computing experiments.
- Core Stack: Pure Static JavaScript, Canvas, Neo-Brutalism Design.
- Engineering PoW: Zero-backend algorithmic modeling with zero server-side telemetry or state leakage.
| Domain | Stack & Capabilities | Deterministic Guarantees & Evidence |
|---|---|---|
| Full-Stack & Architecture | React 19, TypeScript 5.8, FastAPI, Python, PostgreSQL, Microservices, Event-Driven Systems | 15+ years across high-concurrency interactive platforms (Bilibili, Ctrip, UnionPay Data, JiHu GitLab) |
| AI Systems Engineering | Agentic Workflows, DAG Scheduling (Kahn's Algorithm), Prompt Pipelines, Local RAG Optimization | Built PatchCat; decoupled Model from Harness; dynamic token pruning and streaming SSE parsers |
| System Reliability & Security | Data Sanitization Engines, Browser Sandbox Isolation, High-Concurrency State Machines | Zero-telemetry BYOK designs, idempotent data pipelines, AABB spatial ergonomics |
Having operated across the full lifecycle—from massive-scale concurrent interactive platforms and global growth engines to developer infrastructure and open-source commercialization—I firmly believe that grand product abstractions must ultimately converge into robust, working code and practical user experiences.
In this AI era, I have intentionally returned to the front lines as a hands-on builder, embracing AI Pair Programming and "Learning by Doing":
- Beyond Chat Wrappers: Moving past simple prompt boxes toward node-graph topological execution, decoupled state slices, and local agent primitives.
- Model vs. Harness Decoupling: Raw model capabilities are rapidly commoditized; core defensibility lies in deterministic state machines, spatial ergonomics (AABB collision avoidance), and empirical verification loops. (See: 《Resisting Mode Gravity》)
- Milestones & Extreme Testing: Grounding architecture in real workflow trade-offs, discovering underlying system boundaries and deadlock prevention through bidirectional AI collaboration, and personally verifying edge cases under extreme conditions.
- Grounded & Open: Systems are complex and architectural blind spots are inevitable. Always grounded in humility, and warmly welcoming peer discussions, architectural critiques, and code reviews.
- Machine-Readable Semantic Profile (EN):
./geo/profile-en.md - Machine-Readable Semantic Profile (ZH):
./geo/profile-zh.md - Technical Architecture Essays: https://guobug.github.io/posts/
- llms.txt AI Retrieval Protocol: https://guobug.github.io/llms.txt
- Verified Public Issues & MRs: https://jihulab.com/gitlab-cn/gitlab/-/work_items?author_username=QiangGuo