Sirui Li

Bank of China, Shenzhen Branch | Backend Developer and AI Intern

June 2024 – Aug 2024

• Engineered Retrieval-Augmented Generation (RAG) technology with large language models to enhance the bank’s recommendation system, reducing customer service response time by 10-15% and increasing system availability by 10%.
• Created an internal service application that streamlined inquiry processing through LLM integration, achieving a 25-30% reduction in customer service handling time.
• Collaborated cross-functionally to implement backend infrastructure improvements and UI/UX refinements, significantly increasing daily active users and enhancing overall customer satisfaction.

X

Factory Worker Safety Monitoring System

Researcher under Prof. Yin Li | University of Wisconsin-Madison, Department of Computer Science

Sep 2024 - Present

• Developed pose estimation pipelines integrating Intel RealSense with LightBuzz and MediaPipe, optimizing real-time RGB-D data fusion and enhancing computer vision performance.
• Modernized Intel RealSense-based Android application through comprehensive API updates and new motion detection features, achieving 15-20% improved accuracy in industrial safety monitoring.
• Developed rule-based algorithms to detect and analyze worker actions including walking, striking, pushing, and pulling from 3D pose data for GM RFC checklist compliance.
• Contributed on establishing comprehensive video data annotation workflows and datasets creation from GM factory footage for both rule-based systems and machine learning models.

Visual-Language Robotic Control System

Researcher under Prof. Beichen Ding | Sun Yat-sen University (Shenzhen), School of Advanced Manufacturing

May 2024 - Sep 2024

• Integrated advanced computer vision with robotic systems through LoRA-based language models, enabling autonomous real-time decision-making for robotic actions based on visual inputs.
• Developed a visual encoder pipeline combined with large language models for human keypoint detection, reducing computational overhead by 20% while maintaining precision.
• Enhanced robotic response accuracy through optimized human keypoint localization, focusing on precise body part identification, such as left shoulder detection, for improved operational efficiency.

Biomolecular Dynamics Modeling System

Researcher under Prof. Xuhui Huang | University of Wisconsin-Madison, Department of Chemistry

May 2023 - Sep 2023

• Leveraged statistical models for biomolecular dynamics by implementing IGME function in MSMBuilder, utilizing Python and PyTorch to enhance molecular behavior prediction capabilities.
• Applied advanced Markov State Models and Generalized Master Equation methodologies to improve the accuracy and predictive power of biomolecular dynamics modeling.
• Implemented Minimum Variance Cluster Analysis for MSMBuilder2022, significantly improving computational efficiency in molecular state prediction and clustering.