Experience

Industry Research

• 06/2024 - 08/2024: Model Optimization Intern at Biren Tech, Shanghai, China
  • Summary: Large Language Models and Distributed Computing
  • Verified the stability of large models (such as InternLM with 7B, 70B, and 100B parameters) on a 4-node, 32-GPU setup, focusing on training and fine-tuning these models in a distributed system.
  • Optimized the performance of existing large models on BR GPUs by designing parallel strategies and conducting performance analysis on a 128-GPU setup.
  • Conducted preliminary research on optimizing distributed systems for checkpointing and resuming large models, enhancing the efficiency and flexibility of checkpoint saving.
• 09/2023 - 03/2024: Research Intern at Active Intelligence Pty Ltd, Perth, Western Australia (WA)
  • Summary: Research on Video Anomaly Detecion
  • Optimized the existing model so that the model can detect various abnormal conditions (such as fights, car accidents, fires, etc.) from the surveillance video.
  • Presented the Multi-Scenario Anomaly Detection (MSAD) dataset, a high-resolution, real-world anomaly detection dataset encompassing diverse scenarios and anomalies, both human and non-human-related.
  • Used self-supervised learning and few-shot learning methods, the model is trained in different scenarios, and then fine-tuned in specific scenarios, which improves the generalization of the model. The AUC is increased by 6% on the private dataset and 13% on the public dataset.
• 02/2024 - 04/2023: Computer Vision Intern at ANJI Technology, Shanghai, China
  • Summary: Industrial Implementation of Anomaly Detecion Systems
  • Assisted in the development of the intelligent visual security management application, which is mainly used for the anomaly detection of warehouse staff and early warning of security risks.
  • Assisted algorithm engineers in data mining and data slicing, using CenterNet model to obtain abnormal behavior from a large amount of video frames and provide raw data for model training.
  • Independently developed a visual system for detecting the fuses on automobile parts by using OpenCV, to identify whether the fuses at each specific position are installed in place on the image captured by industrial camera.

Academic Research

• 02/2024 - 11/2024: Master’s Research Student
  • School of Computing, ANU College of Engineering, Computing and Cybernetics
  • Supervisor: Dr. Lei Wang (ANU)
  • Research project: Advancing Video Anomaly Detection
  • One paper has been accepted at NeurIPS 2024 Dataset and Benchmark Track
  • Poster presentation at the ANU Student Research Conference
• 11/2023 - 01/2024: Summer Research Student
  • School of Computing, ANU College of Engineering, Computing and Cybernetics
  • Supervisor: Dr. Lei Wang (ANU)
  • Received the Active Intelligence Research Challenge Award
  • Research project: Advancing Video Anomaly Detection: A Concise Review and a New Dataset [paper]
• 02/2023 - 06/2023: Undergraduate Research Thesis
  • School of Communication and Information Engineering, Shanghai University
  • Outstanding Undergraduate Program of Shanghai University
  • Supervisor: Prof. Yongfang Wang (SHU)
  • Research project: A Lightweight DeepFake Detection Method Based on Spatial-temporal landmarks
• 06/2022 - 07/2022: Summer Research Student
  • School of Communication and Information Engineering, Shanghai University
  • Supervisor: Prof. Yongfang Wang (SHU)
  • Research project: Case analysis of Face Recogition Based on DeepFace