Shiwei Pan

Shiwei Pan

B.Eng. Student, Intelligent Manufacturing Engineering
Tongji University

I am a B.Eng. student in Intelligent Manufacturing Engineering at Tongji University, Shanghai. My work focuses on smart manufacturing and industrial AI, with a particular interest in embodied AI for physical tasks, robotics and intelligent systems, learning-based perception, planning, and decision-making, and vision-language models for industrial inspection.

I am especially interested in building deployable intelligent systems that connect perception, reasoning, planning, and control in manufacturing environments. My recent research spans robotic simulation and manipulation for intelligent construction, ROS-based mobile robot navigation, and multimodal inspection agents for industrial defect detection.

Research Interests

  • Smart Manufacturing and Industrial AI
  • Embodied AI for physical tasks
  • Robotics, intelligent systems, and sim-to-real transfer
  • Learning-based perception, planning, and decision-making
  • Vision-language models for industrial inspection and defect localization

Education

  • Sep 2022 - Jun 2026, Tongji University, Shanghai, China
    B.Eng. in Intelligent Manufacturing Engineering, GPA: 4.55/5.0
    National Scholarship Recipient; Outstanding Student Award
    Relevant coursework: Machine Vision, Industrial Big Data Analytics, Artificial Intelligence, Robotics, Intelligent Production Systems, Optimization Methods

Research Experience

Undergraduate Researcher, Intelligent Construction Robotics
Tongji University, Key University Research Project, 2024 - 2026

  • Built a high-fidelity simulation environment in NVIDIA Isaac Sim for autonomous bricklaying, covering scene construction, sensor modeling, and synthetic data generation for perception development and system-level testing.
  • Designed robotic kinematics, trajectory planning, and manipulation workflows for pick-and-place and bricklaying tasks, with attention to sim-to-real transfer challenges.
  • Developed a ROS-based mobile robot navigation pipeline integrating LiDAR sensing, odometry fusion, SLAM, path planning, and MPC-based motion control; validated the pipeline in simulation and physical environments.
  • Integrated perception, mapping, planning, and control into deployable end-to-end embodied workflows, and contributed to technical documentation and publication-oriented research outputs.

Undergraduate Thesis, Multimodal Inspection Agent for Industrial Defect Detection
Tongji University, 2025 - Present

  • Designed a multimodal inspection pipeline integrating task understanding, ROI localization, promptable segmentation, and structured result interpretation using vision-language models.
  • Proposed an LLM-based adaptive routing strategy for segmentation method selection, enabling dynamic switching between text-prompt-driven and ROI-guided approaches for fine-grained defect localization.
  • Implemented an LLM-orchestrated crop-and-zoom inspection loop in which the model calls a cropping tool to magnify suspicious regions for detail verification, achieving an 8% improvement in multimodal defect detection accuracy on the MVTec AD benchmark.

Publications

Robotica 2025
Robotica publication

Efficient Trajectory Planning for a 4-DOF Robotic Arm with Curve Interpolation and Gaussian Process Inference for Pick-and-Place Manipulation Tasks

Shiwei Pan, Jiaxue Li, Xiaoxiao Lv, Wenrui Jin

Robotica, Cambridge University Press, 2025. Published online May 15, 2025.

DOI

  • Developed trajectory planning methods for a 4-DOF robotic arm, combining curve interpolation and Gaussian process inference for pick-and-place manipulation tasks.

Industry Experience

Digital Product Manager, Hesai Technology
Shanghai, China, Jan 2026 - Apr 2026

  • Digitalized the MES equipment inspection lifecycle, including configurable inspection templates, automated task creation and assignment, formula-driven validation against thresholds, and structured review-and-approval flows.
  • Modeled parallel electrical testing routes for the Sharpa dexterous hand in MES, reducing monthly idle waiting time by 15 hours and improving line throughput.
  • Developed an AI-driven CRM automation module for natural-language order, shipment, and delivery-status queries.

Digital Transformation Intern, Maersk
Shanghai, China, Jun 2025 - Dec 2025

  • Developed an AI-driven procurement analysis pipeline using LLM-based workflows to extract structured data from supplier quotations and integrated ChromaDB for semantic retrieval.
  • Applied Total Cost of Ownership modeling for supplier evaluation, contributing to a 15% improvement in procurement decision efficiency.
  • Built automated feedback and analytics workflows in Power Automate and Power BI to support data-driven supply-chain monitoring.

Data Project Manager, Sapien AI
Beijing, China, Dec 2023 - Sep 2024

  • Managed large-scale AI data projects, including MathGPT and autonomous-driving point-cloud annotation.
  • Designed semi-automated annotation pipelines, quality-evaluation frameworks, and a real-time progress monitoring and anomaly detection system, contributing to an 8% improvement in annotation accuracy.

Skills

  • Programming: Python, C++, JavaScript, ROS, MATLAB
  • AI and ML: Deep Learning, Computer Vision, LLMs, Vision-Language Models, Explainable AI, RAG
  • Robotics: Kinematics, Motion Planning, SLAM, MPC, Sim-to-Real Transfer, Embodied AI
  • Simulation: NVIDIA Isaac Sim, MuJoCo, ROS2, Synthetic Data Generation
  • Manufacturing: Smart Manufacturing, MES Systems, Industrial AI, Process Optimization

Awards and Honors

  • National Scholarship Recipient
  • Outstanding Student Award, Tongji University
  • Second Prize, Shanghai Mathematical Modeling Competition
  • Bronze Medal, University Physics Competition (USA)