About
I am currently an undergraduate at NYU Shanghai, majoring in Data Science, with a double major in Business & Fin and a minor in Computer Science. Interested in full-stack development using React + Django. Also a computer vision researcher in Professor Guo Li's team, working primarily within few-shot learning arena.

Coney Island, NY, NY
Full-stack Developer
Computer Vision Researcher
- Degree: B.S. in Data Science, Business & Finance
- School: NYU Shanghai
- Website: https://nigellu.github.io
- Pronouns: he/him/his
- Email: xl3139@nyu.edu / nigellu@outlook.com
- City: Suzhou, China
-
Revised Feb 21, 2023
Experience
Education
B.S. Data Science & Business and Finance
Sep 2019 - Present
NYU Shanghai, Shanghai, China
Cumulative GPA: 3.893
Minor in Computer Science
B.S. Data Science & Business and Finance
Sep 2021 - May 2022
New York University, NY, US
Studyaway Year at NYU CAS & NYU Stern
Activities
Teaching Volunteer
Stepping Stones, China
Sep 2019 - June 2020
Minhang, Shanghai, China
- Teaching English on weekends for a class of 52 elementary-level migrant children
- Prepare each week's course, including designing and making the course materials, communicating with local teachers
Work & Internships
React JS Frontend Intern
eBay Inc.
IE Team, CCoE
Sep 2022 - Present
Pudong New Area, Shanghai, China
- Build and enhance eBay's ATB (Average Time to Business) dashboard for real-time monitoring of cluster status to facilitate smoother rollouts
- Harness the power of function based ReactJS, Antd, and Redux to create smooth, user-friendly frontend pages for eBay's cloud console
- Continuously optimize the frontend codebase and business logic for faster load time, better user experience, and better code readability, including but not limited to componentization of UI parts, reduced number of API calls, detailed documentation following Google JS Style Guide
Vue JS Frontend Engineer
Kaizntree Ltd.
Sep 2022 - Present
Hong Kong SAR
- Established a startup with 3 of my college friends and 4 other employees, helping more than 30 small businesses to simplify their workflow
- Leverage the power of VueJS and Django's versatility to build a one-stop management system for small businesses. The SPA (single-page-application) we built integrates an all-in-one workflow from raw material purchase to building and selling products (visit Kaizntree website here)
Chinese Admissions Ambassador
NYU Shanghai
Oct 2019 - Present
Pudong, Shanghai, China
- Collaborate with different departments in holding events, online panels, presenting weekly campus tours in representation of NYU Shanghai
- Assist in the admission process by performing data analysis on incoming student applications and providing quantitative insights
DevOps Intern
Jiangsu Expsoft Ltd.
May 2021 - Sep 2021
Wuxi, Jiangsu, China
- Collaborated with a team of 18 people in building and operating a browser-server system based on SpringBoot web framework, Apache Tomcat and MS SQL Server
Deep Learning Intern
Jiangsu Expsoft Ltd.
Sep 2021 - Dec 2021
Wuxi, Jiangsu, China
- Built a NLP segmentation tree API specialized in civil and construction engineering setting, and deployed it as a web-based system with UI and API
- Using MobileNetV2 as feature extraction backbone network, build a liveness anti-spoofing detection network, and deploy it as a system with UI and API; Build a face recognition API based on DeepFace library from FaceBook (now changed name to Meta)
Resident Assistant
NYU Shanghai
Aug 2020 - Jun 2021
Pudong, Shanghai, China
- Oversaw a residence hall with 688 residents
- Developed a safe and inclusive community for 49 students; Collaborate with a team of 31 people on implementing programs and administrative responsibilities; Assess and respond to crisis situations
Publication & Preprint
Nuclear Physics B
Multimodal Online Student Engagement (MOSE) Dataset for Studying Personality and EngagementHanan Salam, Saloni Rakholiya, Jialin Li, Nigel Lu (my preferred English name). Multimodal Online Student Engagement (MOSE) Dataset for Studying Personality and Engagement. Submitted to the Nuclear Physics B, November 2022
CVPR 2023
Boosting Few-Shot Segmentation via Instance-Aware Data Augmentation and Local Consensus Guided Cross AttentionLi Guo, Haoming Liu, Chengyu Zhang, Yuxuan Xia, Xiaochen Lu, and Zhenxing Niu. Boosting Few-Shot Segmentation via Instance-Aware Data Augmentation and Local Consensus Guided Cross Attention. Submitted to the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
Research
Few-Shot Segmentation with Adaptive Data Augmentation and Cross Attention
NYU Shanghai
Mar 2022 - Present
- Research Assistant advised by Professor Guo Li
- Propose an instance-aware data augmentation strategy to improve support image diversity and reduce distribution inconsistency between query and support images in low-data regimes
- Incorporate a cross attention module with 4-D dense correlation refined by local consensus constraints to align query and support features for improved generalization ability
- Set up a neat and re-usable visualization code module to help verify and visualize the results of the proposed model. And build a neat and scalable codebase for few-shot segmentation research in PyTorch
- Co-author a paper which has been submitted to the conference of CVPR 2023
Evaluating Parameter-Efficient Tuning Methods in Low-Data Regimes
New York University
Sep 2021 - Dec 2021
- Course Project mentored by Professor Sam Bowman
- Reproduce four SOTA parameter-efficient tuning methods based on HuggingFace and OpenDelta libraries. Then evaluate the performance of these methods on various NLP tasks (e.g., sentiment analysis, Q&A, etc.) with different portions of training samples provided
- Verify and conclude that a parameter-efficient tuning method with a larger ratio of tunable parameters generally results in a better performance across NLP tasks, but usually converges slower regardless of the sufficiency of data
Multimodal Online Student Engagement Dataset
SMART-LAB NYU Abu Dhabi
May 2021 - March 2022
- Research Assistant advised by Professor Hanan Salam
- Build an online learning engagement detection dataset along with post-class survey results and open to research community
- Based on PyTorch, OpenCV and OpenFace, establish a CNN-RNN hybrid baseline model for spatial and temporal processing
- Investigate the correlation between post-class self-evaluation results and personality surveys to provide insights on how personality may relate to engagement level in online learning. Apply various significance tests (e.g., T-test) to verify the potential correlations discovered
- Co-author a paper which has been submitted to the Nuclear Physics B journal in Nov 2022
Real-time Object Detection in Autonomous Driving Scenarios
New York University
Sep 2021 - Dec 2021
- Course Project mentored by Professor Augustin Cosse
- Pre-process and prepare CityScape dataset for object detection under autonomous driving scenarios (e.g., remove classes irrelevant to autonomous driving). Then Fine-tune YOLOv3 model on the prepared dataset: by freezing the feature extraction backbone DarkNet53 and fine-tuning the feature pyramid network to better detect what matters in driving situations
- Achieved a detection frame rate of 42 on personal computer, with a mAP of 49.6% when transferred to testing images outside CityScape dataset
Contact & Links
Work Email:
xl3139@nyu.edu
xl3139@stern.nyu.edu
Personal Email:
nigellu@outlook.com
orangelu0328@gmail.com