麟 黄

At 上海科爱信教育科技有限公司, I defined requirements, designed solutions, and set delivery boundaries for AI capabilities in Kax. I designed a gateway for DeepSeek, Qwen, Gemini, and OpenAI, with cross-provider fallback and output repair strategies.

I took the RAG knowledge base from zero to production by defining its processing pipeline and quality gates. It produced 325 core knowledge entries and 537 low-frequency fallback entries, with checks for coverage, empty entries, real names, and sensitive information.

I designed safety boundaries for the Agent runtime, including path protections, sensitive-file rejection, command restrictions, and PII safeguards. I also established prompt regression evaluation using paired runs on real conversations, with fail-closed release checks.

Earlier, at 上海朝阳永续信息技术股份有限公司, I developed Java backend services for a fund research platform, including Elasticsearch search, document parsing, caching, and asynchronous tasks. I also contributed to a FastGPT-based knowledge-base Q&A system with PDF parsing, OCR, and SSE streaming, and maintained the fund module backend of the Wind financial terminal at 万得信息技术股份有限公司（Wind）.