Java · Architecture · Full-stack Development · AI Applications
Nearly 15 years in software development, including 5 years as a technical team lead. Experience includes Java business systems for telemarketing, messaging, orders and payments, alongside frontend development and contributions to AI applications. Led a team of around 8, with responsibility for task planning, code reviews, technical guidance, hiring and performance reviews.
Responsible for technical design and core development for telemarketing and messaging systems; contributed to AI customer service and operations self-service projects. Led a team of around 8, organizing requirements reviews, development iterations, and code reviews while coordinating delivery with product, QA, and operations.
Responsible for architecture and core development for Yun Xiao Er and Yun Zhang Gui. Led the order module, developed the payment gateway, and contributed to risk control. Delivered order notifications through persistent Netty connections and JPush.
Contributed to frontend development for the Youle Wang, You Sheng Huo, and You E Dai apps. Improved initial loading through code splitting, lazy loading, image compression, and CDN delivery, and addressed list scrolling performance.
Responsible for Alipay external escrow transaction and batch refund modules. Independently designed refund idempotency checks and distributed locking; used transaction compensation and retries to handle inconsistent transaction states.
Maintained and developed frontend features for AstraZeneca's portal, improving page loading through lazy loading and image compression.
Led the migration from a monolith to microservices and introduced XXL-JOB and Apollo. Addressed lock contention, timeouts, and out-of-memory failures through JVM tuning, SQL index and Redis cache optimization, and hot/cold data separation.
Responsible for platform architecture and core development, using Kafka for asynchronous delivery requests. Led the chain-of-responsibility rule engine for blocklists, rate limits, and template checks, and integrated IVR, SMS, push, and bot channels.
The system receives inquiries through RabbitMQ, saves processing records, then calls the model asynchronously. It combines business context and Bailian knowledge retrieval to draft replies, stores references, prompt versions and model-call details, and supports routing to human agents.
The assistant lets operations staff query business data, statuses and rules. Dify orchestrates intent recognition, parameter extraction, internal API calls and responses, with a 360Teams bot as the query interface.