Dalian, Liaoning, China
Research in computer vision and natural language processing fields and follow up the cutting edge achievements. Led the development of an LLM-powered internal knowledge and recommendation system, improving developer efficiency and reducing bugs by proactively delivering relevant insights from historical documentation. Researched LLM architectures and applications; led the development of LLM-based products, including domain-specific chat systems via LoRA fine-tuning and knowledge base solutions using LangChain and LlamaIndex. Led the development of an LLM-based bug report analysis system that distilled insights from historical reports to support active development, improving efficiency and reducing costs. Led the development of an AI-powered multimodal testing system for in-vehicle platforms, leveraging LLMs and multimodal models to enable full-chain automated testing via AI agents—significantly reducing manual testing costs.
Research in computer vision field, follow up the cutting edge achievements mainly on human face direction and data mining in high dimensional space. Build, deliver and maintain an end-to-end Face Recognition System(FRS) in unconstrained environment; The system consists of several modules including face detection, quality assessment, recognition, clustering with GCNs, etc. Work in other face-related directions including high resolution face image generation with GANs, 3D face reconstruction with 3DMM, multi-pose image generation, etc. Take the responsibility of human face group’s daily work, including work assignment, progress tracking, tech sharing, program discussion, reporting to superior, etc. Participate in open source projects, make contributions in face clustering(data mining) projects.
Research in computer vision field, follow up the cutting edge achievements mainly on generic object detection and image generation. Design and developed an Text Detection-Recognition system for Urban Water Management Business. Design and developed an object detection system based on MaskRCNN, the system is used for detecting and masking out the logo part on cloths.