AI R&D Tech Lead
**** Information Technology
Nov 2024 - Present
- Architected a full RAG pipeline from scratch with hybrid BM25 + vector retrieval and Cross-Encoder reranking, achieving 90%+ retrieval accuracy across million-scale documents.
- Built an LLM-powered query rewriting module, boosting complex query recall by 40% in multi-turn conversation scenarios.
- Designed efficient Qdrant vector indexing with metadata filtering, supporting semantic search at enterprise scale.
- Delivered high-concurrency streaming APIs on FastAPI with Redis caching, cutting average response latency by 30%+.
FastAPI
Langchain
LlamaIndex
VLLM
Qdrant
Redis
MySQL
Docker
K8s
Linux
GitAI R&D Tech Lead
**** Information Technology
May 2024 - Dec 2025
- Designed a three-layer AI architecture orchestrating 5 teams and 12 Agents for complex task decomposition and multi-role collaborative reasoning.
- Implemented multi-Agent scheduling with LangChain and CrewAI, featuring CoT + Tool-calling reasoning flows and shared context management.
- Built a RAG system fusing industry knowledge bases with real-time policy data via Chroma, enabling semantic-level retrieval and generation.
- Achieved 95%+ report automation rate, reducing generation time from hours to minutes and boosting business conversion by 30%.
Langchain
CrewAIFastAPIVLLM
ChromaRedisMySQLDockerK8sLinuxGitSenior Game Server Developer
**** Games
Apr 2022 - May 2024
- Designed and implemented core gameplay APIs and server-side state management for multiple overseas game titles.
- Built game economy balancing systems with statistical modeling, ensuring long-term numerical stability.
- Developed an anti-cheat detection system with a Django-based admin dashboard for real-time monitoring and policy management.
- Optimized high-concurrency backend with Python Asyncio and Redis caching, improving API response performance by 20%+.
Python
Golang
DjangoRedisMySQLDockerLinux
NginXGitPython Backend Developer
**** Technology
Jul 2019 - Apr 2022
- Built high-performance backend services on Tornado, powering article management, scheduled publishing, and content moderation modules.
- Designed a template-driven image generation service handling 80M+ monthly calls with sub-300ms average latency.
- Engineered Redis caching strategies for assets and intermediate data, significantly reducing I/O overhead and processing time.
- Developed a Scrapy-based scraping system for automated content collection with anti-crawl countermeasures.
PythonDjangoFastAPIRedisMySQL
MongoDBDockerLinuxNginXGitContact Form
Please contact me directly at supermannb999(at)gmail.com or drop your info here.