Curriculum Vitae
Download my most recent CV here (updated: Dec. 2024).
Education
Carnegie Mellon University, School of Computer Science
Master of Science in Machine Learning | 2024/01 - 2024/12
- QPA: 4.16 / 4.0
- Major Courses: Advanced NLP, Probabilistic Graphical Models, Advanced CV, Deep Learning Systems
Bachelor of Science in Artificial Intelligence | 2020/09 - 2023/12
- QPA: 4.0 / 4.0
- Additional major in Psychology
- Major Courses: Algorithm Design & Analysis, Operating System, Multimodal Machine Learning, Probability Theory, Modern Regression
Work Experience
Software Development Intern | 2024/05 - 2024/08
- Developed a new image editing panel using C++ and Qt, enabling users to annotate, rotate, crop, and adjust image colors efficiently.
- Collaborated with developers, UI designers, testers, and business teams to refine control logic, UI, and data management based on feedbacks.
- Presented a working prototype to the development team; delivered a final presentation to the entire company upon product update release.
Quant Trader Intern | 2023/06 - 2023/08
- Analyzed six months of trade data to assess strategy performance during NASDAQ closing auctions; identified key factors affecting execution cost.
- Provided daily updates to mentor and weekly reports to manager; delivered a final presentation with data visualizations to enhance trading insights.
Software Development Intern | 2022/06 - 2022/08
- Worked with UI/UX teams to develop logic and layouts for a weight-balancing game, improving gameplay and user experience.
- Optimized Java-based UI engine, isolated core functions from game logic, improved testing efficiency by 40x for faster development cycles.
Zhengren Quantitative Investment
Strategic Algorithm Intern | 2020/12 - 2021/01
- Implemented a new strategy to reduce transaction cost; analyzed the stability of the outputs of prediction models.
- Built a back-testing system upon the data interface; gave presentations on the performance of different strategies.
Selected Projects
- Developed iterative vessel segmentation using Hessian features; reduced intensity variation by 35%, improving accuracy of vessel analysis.
- Published 3 abstracts and presented at OHBM and AAIC regarding the method and statistical analysis on the features of vascular structures.
- Integrating segmentation data to train a transformer-based 3D U-Net for more robust results and faster inference.
- Implemented a UNIX-inspired preemptive kernel on x86 architecture from scratch that provides a system call interface to the user.
- Developed a Pthreads-like thread library and a thread management interface that allows users to run multithreaded C programs.
- Collaborated with a teammate, met twice a week to update progress, coordinate tasks, solve problems and divide work; finished the project on time.
- Improved LLM performance on WebArena subset by 10% with self-generated critiques on trajectory, enhancing long-term reasoning abilities
- Conducted ablation studies that found the ability of critic model is more effective to improve the performance of self-reflection agent.
- Developed an end-to-end retrieval augmented generation system on information about CMU from scratch.
- Collected raw data from websites, open access papers, and plain text * documents; compiled knowledge resource by reasonable segmentation.
- Developed a complete pipeline with document retrieval with ColBERT and * generation with zero-shot and few-shot prompting T5 and Llama 2.
- Created training set with Self-Instruct pipeline; annotated validation set and evaluated validity.
- Designed and developed a question-answer generation pipeline based on CLEVR for new question types regarding relationships between objects.
- Co-authored and published the free-form exploration benchmark EXCALIBUR: Evaluating and Encouraging Embodied Exploration in CVPR 2023.
- Introduced more vision-language interaction into SoTA model on ALFRED dataset; increased task success rate by 2.8% (seen) and 2.5% (unseen).
- Trained a ViLT and a BLIP to integrate the instructions into the semantic map generation module; decrease KL divergence from 0.62 to 0.59.
- Few-shot prompted GPT-4 and fine-tuned ada to improve subtask decomposition accuracy from 80% to 92%.
- Developed a demo that allows users to place AR spatial anchors, mark accessible regions, and localize within accessible regions.
- Provided API to a question answering system that provides real-time navigation instructions based on localization result.
Skills
Programming Languages
- Proficient: Python, C++, R, C, x86 Assembly, Java
- Elementary: Swift, Standard ML
Programming Libraries
- Python: NumPy, PyTorch, pandas, OpenCV, scikit-learn, Tensorflow, spaCy, lightgbm
- C++: Qt, Insight Toolkit (ITK), Visualization Toolkit (VTK)
- LaTeX, Git, Docker, Markdown, Unix, Vim
Languages
- Native: Chinese, English
- Elementary: Japanese
Publications
Shi, Z., Li, J., Huang, Y., Schweitzer, N., Iordanova, B., Ibrahim, T. S., Yang, S., Stetten, G. D., Aizenstein, H. J., Wu, M. (2024, June 24-25). Exploring Age-Related Morphological Changes in Cerebral Arteries: A 7T TOF MRA Study [Poster Abstract]. The Organization for Human Brain Mapping 2024 Annual Meeting, virtual.
Li, J., Stetten, G. D., Schweitzer, N., Shi, Z., Ibrahim, T. S., Yang, S., Iordanova, B., Aizenstein, H. J., Wu, M. (2023, July 15). Age- and Sex-Related Morphological Changes in Cerebral Blood Vessels: a 7T TOF MRA Study [Poster Abstract]. The Alzheimer's Association International Conference 2023, virtual.
Li, J., Stetten, G. D., Schweitzer, N., Ibrahim, T. S., Iordanova, B., Aizenstein, H. J., Wu, M. (2023, June 23-24). VesselMapper: A Robust Vessel Segmentation Algorithm for 3D Images [Poster Abstract]. The Organization for Human Brain Mapping 2023 Annual Meeting, virtual.
Zhu, H., Kapoor, R., Min, S. Y., Han, W., Li, J., Geng, K., ... & Weihs, L. (2023). Excalibur: Encouraging and evaluating embodied exploration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 14931-14942).
Teaching Experience
Carnegie Mellon University
21-127: Concepts of Mathematics
