Resume
Basics
Name | Shengjie Li |
sxl180006@hlt.utdallas.edu | |
Phone | (732) 485-7674 |
Url | https://shengjie-li.com |
Education
-
Aug. 2020 - May 2025 University of Texas at Dallas
Ph.D. in Computer Science
-
Sep. 2018 - May 2020 Rutgers University
M.S. in Computer Science
-
Sep. 2013 - June 2017 South China Agricultural University
B.Eng. in Software Engineering
Work
- May 2023 - Aug. 2023
Applied Scientist Intern
Amazon
- Researched and developed an LLM-based workflow to auto-extract customer-centric product metadata for 5 types of products on amazon.com, poised to significantly automate labor-intensive tasks.
- Augmented LLMs with external knowledge sources, ensuring evidence-grounded results and reducing the risk of inaccurate model outputs (hallucinations).
- May 2021 - Present
Research Assistant
Human Language Technology Research Institute at UT Dallas
- Designed a model that automatically extracts and filters syntactic and semantic features from student essays, achieving a 1.6% improvement in cross-prompt essay scoring over previous state-of-the-art models.
- Built a pipelined model for entity coreference resolution, achieving the best performance in the CODI-CRAC 2022 shared task which is 1.6x as good as the baseline and 1.2x as good as the second-ranked team.
- Developed a top-performing end-to-end model for the discourse deixis track in the CODI-CRAC 2021 shared task, utilizing resolution constraints to achieve a performance of 1.8x as good as the second-ranked team.
- Developed an end-to-end model for discourse deixis resolution that leveraged task-specific characteristics and outperformed previous state-of-the-art by 27%, resulting in a first-author publication in EMNLP 2022.
- Collected and analyzed data for identifying propaganda content in Spanish magazines during World War II. Identified key challenges for developing a deep learning model. Published findings in AAAI 2023.
- June 2019 - May 2020
Research Assistant
NLP Group at Rutgers University
- Developed a multimodal classification model to predict the coherence relation between an image and its caption, aiding the development of controllable caption generation and leading to an ACL 2020 publication.
Service
- Conference Reviewer: EMNLP 2023, ACL 2023, ACL ARR
- Journal Reviewer: AIJ, TALLIP
Awards
- 2022--2024
Louis Beecherl, Jr. Graduate Fellowship at UT Dallas
Jonsson School at UT Dallas
Merit-based fellowship for graduate students with a competitive academic record.
- 2015
Silver Medal in the 2015 ACM-ICPC (International Collegiate Programming Contest) Asia Shenyang Regional Contest
Northeastern University (China)
Languages
Cantonese Chinese | |
Native |
Mandarin Chinese | |
Native |
English | |
Fluent |
Japanese | |
Elementary |