Zhang Jia
- 18292060445 · plus_zzz@163.com · 2002-09
- Homepage: pluszzz.top
- GitHub: github.com/Pluszzz
Education
- "Huawei Cup" Graduate Math Modeling — National Second Prize
- National Math Modeling Contest — Provincial Second Prize
- CET-4
Projects
Overview: Enterprise RAG intelligent agent platform covering the full pipeline from document ingestion to intelligent Q&A — intent recognition, question rewriting, multi-channel retrieval, session memory, model routing with fault tolerance.
Stack: SpringBoot + MyBatis Plus + Milvus + Redis + Redisson + Apache Tika + Sa-Token
Responsibilities:
- Multi-Channel Retrieval: Designed multi-channel retrieval engine (intent-directed + global vector + post-processing pipeline)
- Question Optimization: LLM-based question rewriting and decomposition for automatic context completion
- Session Memory: Compression strategy (sliding window + auto-summary) to reduce token consumption
- Model Routing: Three-state circuit breaker for multi-model priority scheduling and automatic failover
- Intent Recognition: Tree-structured multi-level intent classification with confidence-based clarification
- Tool Framework: MCP tool integration with automatic registry-based discovery
- Ingestion Pipeline: Pluggable document processing pipeline via unified interface
- Rate Limiting: Redis ZSET + Lua sliding window for global and user-level concurrency control
- Context Propagation: TransmittableThreadLocal for cross-thread trace context in async thread pools
Overview: SaaS short link system providing efficient, secure URL management with analytics for enterprise and individual users.
Stack: Spring Boot + Spring Cloud Alibaba + MySQL + Redis + RocketMQ + ShardingSphere + Sentinel
Responsibilities:
- MQ Idempotency: Redis-based exactly-once message consumption within time windows
- Cache Protection: Double-check locking with null-value caching for penetration protection
- Bloom Filter: Redisson Bloom filter for efficient short link existence checks
- Rate Limiting: Sentinel QPS-level rate limiting with graceful degradation
- Async Peak-Shaving: RocketMQ async consumption for high-concurrency monitoring writes
- Cache Consistency: "Update DB, delete cache" strategy for data consistency
- Distributed Locks: Redisson read-write locks for concurrent data modification
- Sharding: ShardingSphere shard routing for targeted queries
Skills
AI Coding
AI Agent development workflows with Codex, Claude Code for code generation, debugging, and automated development
Java & JVM
Java concurrency (JUC, AQS, CAS), thread pool tuning. JVM memory model, GC (G1/CMS), OOM diagnosis, performance tuning
Backend Frameworks
Spring Boot / Spring Cloud Alibaba microservices. Nacos, OpenFeign, Sentinel for service discovery and traffic governance
Databases
MySQL index design, execution plan analysis, MVCC, transaction isolation. Slow query optimization, read-write separation
Caching & Middleware
Redis data structures and persistence. Cache solutions. Distributed locks, sliding window rate limiting, Bloom filters
LLM Applications (RAG)
Enterprise RAG: Milvus vector retrieval, multi-channel recall, question rewriting, session memory compression, model routing
Engineering
Git version control, Maven project builds, modular development