I am a master student of MBZUAI majored in computer vision, as of Aug. 2024, with Prof. Xiaodan Liang as my supervisor. My research areas include cumputer vision, natural language processing and large multi-modal model.

From 2024 August to present, I am a research intern at Foundation Model group in MEGVII Technology, supervised by Yingfei Liu, Tiancai Wang, and Xiangyu Zhang

Earlier I was a research engineer at Megvii Research, supervised by Xinyu Zhou, now co-founder of moonshot ai.

I received my bachelor degree from University of International Business and Economics in 2022, where I spent four wonderful years.

Research Engineer / Megvii Research

Jul 2022 - Jul 2024

Megvii operates the world’s largest computer vision research institute. At Megvii Research, my main work is to do research on large vision-language model and open vocabulary detection.

In addition, I am also responsible for the development of advanced AI algorithm production systems - AI Service, an AI infrastructure designed to adapt to algorithm mass production.

Research Assistant / Tsinghua University(supervised by Prof. Xinyu Zhang)

Aug 2021 - Jul 2022

Contributed to a published paper: "OpenMPD: An Open Multimodal Perception Dataset for Autonomous Driving"

Proposed an autonomous driving dataset, which includes all the mainstream task annotations, such as 2D and 3D bounding box and segmentation, which plays an important role in multi-task and multi-modal fusion network.

Research Assistant / Westlake AI Innovation Lab(supervised by Prof. Stan Z. Li)

May 2023 - Sep 2023

Contributed to a paper: Protein 3D Graph Structure Learning for Robust Structure-based Protein Property Prediction

A protein structure embedding alignment optimization framework is proposed to alleviate the structure embedding bias.

Research Assistant / National Sun Yat-sen University(supervised by Prof. Xiaodan Liang)

Jul 2023 - Dec 2023

Enhanced the BARON baseline by applying self-distillation to augment data, employing teacher models trained for a substantial number of iterations (e.g., 10,000) to label images with novel class tags based on confidence thresholds.

Achieved a notable increase in the model's MAP50 score on novel classes, raising it from 33.2 to 38.2.

Multimodal Large Model Researcher / Megvii Research

Aug 2024 - Present

Focused on multi-modal large model related research topics.

Perception Algorithm Engineer / Nio Inc.

Oct 2021 - Jan 2022

Focused on implementing and optimizing 2D and 3D object detection algorithms for parking perception systems.

Deep Learning Algorithm Engneer / Didi Global Inc.

Aug 2021 - Oct 2021

Focused on developing and implementing security products, including anomaly detection and anti-crawler measures.

Algorithm Engneer / Naxions.

Dec 2020 - Mar 2021

Utilized bio-inspired algorithms for optimizing hospital visit sequences for medical representatives at Naxions.