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 physical AI for manufacturing, autonomous mobile robots, multi-robot coordination, learning-based perception, planning, and control, 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 work spans robotic simulation and manipulation for intelligent construction, ROS-based mobile robot navigation, multimodal inspection agents for industrial defect detection, and AMR algorithms for warehouse robotics.

Research Interests

  • Physical AI for manufacturing
  • Autonomous mobile robots and multi-robot coordination
  • Robotics, intelligent systems, and sim-to-real validation
  • Learning-based perception, planning, and control
  • 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

AI Algorithm Engineer, Quicktron Robotics — AMR Full-Stack Algorithms
Shanghai, China, Jun 2026 - Present

  • Develop order batching and wave-planning algorithms for warehouse AMR systems, connecting business orders to executable task groups, priorities, and robot assignments.
  • Model task dependencies and execution links and contribute to the Task-to-Path pipeline, where global routes and waypoint-level constraints are updated from runtime system state.
  • Work on perception-informed path adaptation, including collision-risk evaluation, local path correction, and escalation to global replanning when routes become blocked.
  • Expanding responsibilities toward LiDAR perception, localization, motion planning, control, and multi-robot coordination as part of end-to-end AMR algorithm development.

Product Manager, Hesai Technology — AI and Manufacturing Digitalization
Shanghai, China, Jan 2026 - Present

  • Developed a conversational AI CRM assistant with Cursor and Playwright MCP, rapidly integrating internal interfaces to support natural-language queries on order status, shipment progress, and estimated delivery dates.
  • Conducted on-site research at Shanghai and Hangzhou factories and led product design and process redesign for visitor management, equipment inspection, and continuous-improvement systems.
  • Built the equipment-inspection workflow from template configuration and task generation through formula-based validation, review, and completion; also modeled parallel PIP/DIP test routes that reduced monthly idle waiting time by 15 hours.

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

  • Built an AI quotation-analysis system that extracted key fields from unstructured documents, used ChromaDB for supplier-quotation retrieval, and applied TCO models for standardized comparison, improving procurement decision efficiency by 15%.
  • Developed an end-to-end supplier management platform with Power Automate for feedback collection, AI-based synthesis of supplier and project-manager feedback, and Power BI dashboards for procurement progress and supplier performance.

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++, ROS/ROS2, MATLAB, JavaScript
  • Robotics: LiDAR Perception, SLAM, Localization, Motion Planning, MPC, Multi-Robot Coordination
  • AI and Simulation: Computer Vision, LLM/VLM Agents, Deep Learning, NVIDIA Isaac Sim, MuJoCo
  • Manufacturing: AMR Systems, Order Batching, Smart Manufacturing, MES, 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)