AI Full-Stack Intern · Baidu (China) Co., Ltd. Guangzhou Branch 2026.07 – 2026.08

Led the security-hardening rewrite of the e.baidu backend system.

• Designed and led a strangler-fig migration from a ThinkCMF 6 + ThinkPHP 6 PHP monolith to NestJS 10 + TypeScript, fixing P0 vulnerabilities (arbitrary SQL execution via WebSocket, hardcoded credentials, superuser hard-bypass)
• Implemented zero-trust JWT gateway integration (UUAP), dual-guard RBAC with strict-deny defaults, and hashed sensitive-field indexing
• Migrated 30 read-only API paths / 39 operations with 85 Jest tests passing; shipped a React 18 + Ant Design Pro admin frontend with an SSE-streamed agent chat preview (13 Vitest tests passing)

AI/LLM Engineering Intern · Guangzhou Xin'an Data Co., Ltd. 2026.04 – 2026.07

Built Multi-Agent document review pipelines for enterprise brand & KPI compliance systems.

• Orchestrated multi-node Agent workflows with Dify, converting manual report review into a fully automated pipeline across 7 intent-routed review dimensions
• Optimized RAG retrieval (hierarchical Markdown chunking + Hybrid Search + Reranker on Milvus) — QA accuracy 60% → 88%
• Built a provider-agnostic LLM connection pool with exponential backoff, token-bucket rate limiting, and streaming support

Founder / Full-Stack Developer · Guangzhou BoBaCarry Smart Tech Co., Ltd. 2025.09 – 2026.04

Built a shared smart-follow-cart platform end-to-end, from WeChat Mini Program to production deployment.

• Node.js + Express backend with MySQL + Redis (<50ms response latency); Vue 3 + Element Plus ops console
• Defined MQTT device protocol with hardware engineers; built a device simulator to unblock firmware integration before hardware arrived

Contributing to the Vercel AI SDK ecosystem (vercel/ai, vercel/ai-elements) — CSS layout fixes, TypeScript API design, and batch-processing algorithm design for the core embedMany function.

HUM / Voices — Emotional voice-persona platform 2026.01 – 2026.02

Winner, AI³ Growth Journey Season 8. Each Voice = timbre (StepFun API cloning) + personality + emotional-boundary strategy.

• SSE-streamed TTS with React state + Audio Context double-buffering — <300ms time-to-first-audio
• Redis semantic/audio caching cut repeated-dialogue TTS cost by 45%
• On-chain persona provenance on 0G (AI + Data native chain)

TypeScript React Node.js PostgreSQL Redis Docker StepFun API

EdTech Platform — Adaptive learning path generation 2025.11 – 2026.01

• Extended BKT cognitive diagnosis model with Redis-backed real-time mastery-probability updates
• Few-shot-injected BKT mastery state into Gemini API for difficulty-calibrated question generation

Spring Boot Gemini API React RabbitMQ MyBatis Plus

indicator_detector — KPI quality detection & recommendation 2026.04 – 2026.05

• Hybrid scalar-filter + vector-similarity retrieval on Milvus Lite for multi-tenant Top-K KPI recommendation
• ReAct-style prompt design (Few-shot + CoT) for formula-dependency parsing with self-correcting JSON Schema output

Python Milvus Lite Transformers PyTorch

Multi-Agent Orchestration (Dify) RAG (Hybrid Search + Rerank) Prompt Engineering (CoT/Few-shot) LLM Connection Pooling JWT/RBAC Zero-Trust Gateway