Yibo Zhao +1-408-***-****

Master of Computer Science # ad4tun@r.postjobfree.com

Department of Computer and Information Science Ñ 479 Kiely Blvd, Santa Clara, CA 95051 University of Pennsylvania U.S. Permanent Resident Skills and Courses

Programming and Skills: Python, Java, C, PyTorch, TensorFlow, OpenAI API, Pandas, Numpy, Sklearn Tools: LLM, RAG, Github, Linux, MakeFile, Docker, GDB, JUnit, Latex, Jupyter notebook Courses: Natural Language Processing (NLP), Computer System Programming, Introduction to Computer Systems, Data Structures and Software Design, Introduction to Software Development Education

•University of Pennsylvania 01/2023-Expected graduation: 05/2025 Master of Computer Science, GPA: 3.80, Admissions Ambassador Philadelphia, PA, U.S.

•Humboldt University of Berlin and Tongji University 09/2015-11/2019 Masters of Law, GPA: 3.90 and 3.60 Berlin (Germany) and Shanghai (China)

•Beijing International Studies University 09/2011-07/2015 Bachelor of Art, GPA: 3.80 Beijing, China

Work Experience

•OpenAI and Remotasks Inc. 02/2024-present

Coding specialist contractor, LLM trainer San Francisco, CA

– Collaborated with OpenAI’s ChatGPT team to enhance the model capability on general coding and programming questions. This project is performed on OpenAI’s Feather platform.

– Delivered high-quality training data for LLM fine-tuning, including user prompts, ideal responses, and auxiliary data for LLM’s intermediate reasoning steps.

– Specific techniques include prompt engineering, response rubric construction, and model evaluation.

•Shanghai Securities Co.,Ltd. 07/2019-12/2021

In-house Attorney Shanghai, China

– Oversaw legal aspects of the headquarters’ wealth management and investment operations, along with securities brokerage across 24 branches, ensuring strict adherence to evolving regulatory standards. Project Experience

•Fine-Tuning GPT-3 for Text Simplification 11/2023 - 12/2023 Technology stack: OpenAI API, Sklearn, fine tuning, prompt engineering

– Leveraged GPT-3 to perform text simplification tasks, applying zero-shot prompting, few-shot prompting, and chain-of-thought (CoT) prompting, creating fine-tuned models from OpenAI’s davinci, currie, and ada.

– Applied tools such as Textstat and Sklearn to measure text complexity and evaluate the quality of the output text.

•Model Training for Text Classification 09/2023 - 10/2023 Technology stack: PyTorch, Python, Hugging Face

– Prototyped two classic language models: N-gram, and character-RNN from scratch using PyTorch.

– Pre-trained Bert (from Hugging Face) using data from Wikipedia, Shakespeare, and web posts (with AutoTok- enizer).

– Performed hyperparameter tuning to improve the quality of text classification tasks.

•Academic Management and Data Ingestion System 06/2023 - 07/2023 Technology stack: Java and JUnit

– Prototyped a course management system for tasks managing among courses, students, and professors.

– Developed data loading and cleaning functions for various data formats (e.g., CSV, JSON, TXT).

– Leveraged critical data structures, including HashMaps, ArrayLists, and Sets, and applied polymorphism to ease the system implementation and improve extensibility.

– Built extensive tests using JUnit.

Contact this candidate