Enterprise Agent Platform Architect
Profile
10 years across large-scale platforms and AI engineering at Alibaba and SenseTime, now leading enterprise Agent platform architecture, productization, and delivery.
- Service scale
- 2k+ tenants / 230k+ users
- Peak traffic
- 60k / day
- Team scope
- 20-person product-engineering / 3-person Agent squad
- Harness coverage
- 40 repositories
Experience
Owned product architecture and production operations for a cross-industry WhatsApp business-Agent SaaS serving 2k+ tenants, 230k+ users, and 60k peak requests/day across RAG/Memory, tools and Skills, quality, cost, and team delivery.
Evolved WhatsApp customer service from rule flows to tenant-configurable LLM business Agents across intent routing, SOPs, knowledge, tools, callbacks, and handoff; delivered 70%+ auto-resolution, about 7s end-to-end latency at 7k+ tokens per conversation, and 60% lower token use.
Built multi-tenant Agentic RAG for documents, QA, websites, attachments, and product data, with ingestion, retrieval, reranking, and an unknown-question/content-gap flywheel; 50% of clusters entered production optimization, high-frequency coverage reached 70%, and auto-resolution improved about 30%.
Built a LangChain DAG-based attentive reasoning Agent that decomposed workflows into intent, rule, tool, parameter, and compliance decisions, with explicit dependencies, termination, fallbacks, and replayability.
Built tenant-isolated session and user Memory for fact extraction, confidence filtering, write/update, and retrieval injection, carrying prior conclusions, preferences, and task state across interactions.
Led an internal AI collaboration CLI and team-wide shared context for releases, operations automation, case diagnosis, project organization, and checks; integrated Skills, MCP, CLI, repository standards, and Spec/EPIC workflows across 40 repositories.
Built and delivered independent AI data and model-serving products, owning product definition, system architecture, full-stack implementation, containerized deployment, and production iteration.
Built a production multimodal data and model platform spanning annotation, model artifacts, auto-labeling, quality validation, and pluggable Nuclio/Traefik inference for audio, video, and image workflows.
Owned web engineering and microservice architecture for AI image generation, model training, and platform products, contributing to 0-to-1 generative AI SaaS delivery.
Led service architecture for proprietary image models, LoRA, and open-source integration, organizing Stable Diffusion/PyTorch inference, queues, model files, post-processing, and observability into asynchronous Web, H5, and mini-program delivery.
Worked across Alibaba Local Services commercial promotion and Taobao new-retail engineering on high-traffic cross-platform systems spanning ad delivery and attribution, live interaction, micro-frontends, SSR, and performance.
Moved Taobao Live interaction from isolated Weex components to H5 micro-frontends with unified routing and communication; combined SSR and tiered resources to keep FCP below 1.5s under high traffic and weak networks.
Owned financial reporting, Java web services, and frontend migration from AngularJS to Vue with a shared Web UI component library.
Built Spring Boot/MyBatis financial operations services, migrated AngularJS to Vue, and established a shared Web UI library.
Skills
Agent Runtime / Agentic RAG / Memory / Multi-agent Orchestration / Workflow Orchestration / Tool-use / Function Calling / MCP / Skills / Human-in-the-loop
Multi-tenant Isolation / Permission Model / Tool Registry / Audit Trail / Guardrails / Replay Evaluation / Golden Set / Trace Observability / Usage / Cost Governance
Python / FastAPI / TypeScript / Node.js / Rust / Java / Linux / PostgreSQL / Redis / Milvus / Docker / Kubernetes / Microservices / CI/CD / Alibaba Cloud / Google Cloud
Model APIs / Routing / Context Engineering / RAG Retrieval / Reranking / PyTorch / Stable Diffusion / Gemini / Data Feedback Loops / LLMOps
Platform Roadmap / Architecture Rules / Spec / EPIC Workflow / Release Gates / Cross-functional Delivery / AI Development Harness / Team Standards