Runtime and Orchestration
Agentic RAG, parallel multi-agent orchestration, unified context, and dynamic Skill routing, with about 7s end-to-end latency under 7k+ average conversation-token load.
Platform Focus
AI Agent Platform Engineering from Architecture to Production Operations
Owning AI Agent platform architecture, tools and permissions, replay evaluation, quality operations, and enterprise integration to bring LLM capabilities reliably into core workflows.
PLATFORM OWNER
Platform Ownership
End-to-end ownership across architecture evolution, core runtime, shared platform capabilities, production quality operations, and team delivery, accountable for reliability, efficiency, cost, and business outcomes.
Platform capability map- 01
Led the evolution from code-guided Agent 1.0 to model-driven 2.0, defining the boundary between model decisions and Runtime control.
- 02
Built Agentic RAG, parallel multi-agent orchestration, unified context, dynamic Skill routing, and MCP/API tool systems.
- 03
Established multi-tenant isolation, tool permissions, audit, human handoff, idempotent retries, and high-risk action controls.
- 04
Connected production traces, replay evaluation, release gates, and data feedback to turn production issues into knowledge, Skill, and policy improvements.
- 05
Led the AI Agent squad in core platform delivery and expanded the AI development harness across 40 repositories, connecting development, testing, release, deployment, and verification.
Core Operating Areas
Connecting architecture decisions, platform capabilities, and team delivery through production outcomes.
Tools and Production Governance
MCP/API, tool gating, multi-tenant permissions, audit, side-effect controls, human handoff, and an observable execution ledger.
Quality and Cost Operations
Continuous optimization through replay evaluation, content gaps, and usage auditing, reducing token consumption by 60% and improving auto-resolution by about 30%.
Team and Engineering Effectiveness
Led a 4-person AI Agent squad within a 20-person product-engineering team; expanded the AI development harness across 40 repositories and reached roughly 10x the prior collaboration model's efficiency for standardized delivery.
Selected Delivery
AI Agent platforms, model serving, and large-scale platform engineering.
2025-03 - Present
Read more Production-grade AI Agent Platform Architecture
Owned a cross-industry AI Agent customer-service SaaS platform from architecture evolution through production operations and expanded the team AI development harness across 40 repositories.
2021-09 - 2024-05
Read more AIGC SaaS and Model-serving Platform
Owned backend engineering, microservice decomposition, inference call chains, and platform operations for 0-to-1 generative AI SaaS delivery.
2017-10 - 2021-08
Read more Alibaba High-traffic Commerce and Live Engineering
Owned high-traffic cross-platform web engineering, micro-frontends, ad attribution, logging SDKs, JsBridge, and performance optimization across Alibaba Taobao and Local Services.
Data Feedback
Read more AI Agent Platform Data Feedback System
Connects production traces, issue cases, evaluation samples, and preference data so online behavior becomes a continuous quality improvement mechanism.
Evaluation Practice
Turning production traces into replay evaluation and release gates.
Agent Replay Evaluation Harness
Built reproducible Agent regression evaluation from production traces for prompt, tool, and model changes.
Read more