Avatar

Yuhai Li

Ph.D. student of Department of Civil and Environmental Engineering

University of California, Los Angeles

Biography

Welcome to my homepage! I am a 5th-year Ph.D. student co-advised by Dr. Ertugrul Taciroglu and Dr. Puneet Gupta. My research interests include data-driven material analysis and applied machine/deep learning in material science.

I am a conscientious student capable of applying machine learning in an industrial setting. I have sufficient knowledge and experience of machine learning libraries such as JAX, TensorFlow, and PyTorch, as well as software development in Java, Python, and C/C++. I possess practical expertise creating optimization and machine learning methods, such as gradient-based optimization, simulated annealing, Monte Carlo methods, Bayesian models, and reinforcement learning, as well as conducting tests on real systems. I am zealous in my pursuit of new abilities.

I was born in Tianjin, China and migrated with my family to the Bay Area, California in 2012. After graduating from De Anza College, I moved to UCLA to pursue a degree in Computer Science. I earned my B.S. in June 2019 and then enrolled in my current Ph.D. program. I am currently residing in Glendale, California, with my perfect wife, Lia.

Besides doing research and coding, I also love sports, traveling and food.

Interests

  • Machine/Deep Learning
  • Material Science
  • Software Development

Education

  • PhD in Civil Engineering, 2024 (Expected)

    University of California, Los Angeles

  • M.S. in Computer Science, 2021

    University of Illinois, Urbana-Champaign

  • B.S. in Compuster Science, 2019

    University of California, Los Angeles

Skills

Software Development

Machine/Deep Learning

Cloud

Database

Java

Python

Experience

 
 
 
 
 

AI Resident

X, the Moonshot Factory

Sep 2024 – Present Mountain View, California
Confidential 0-1 LLM products development
 
 
 
 
 

Machine Learning Engineer Intern

Unnatural Products

Sep 2024 – Oct 2024 Santa Cruz, California
  • Developed a Python library called virtual lab, a custom workflow that facilitates the creation of virtual configurations of motifs from monomers.
  • Enabled users to select source monomers and build theoretically sound motifs, supporting experimental planning and development.
  • Designed a user-friendly interface for seamless configuration, providing flexibility in designing unique chemical workflows.
 
 
 
 
 

Machine Learning Research Engineer Intern

Scale AI

Jun 2024 – Sep 2024 San Francisco, California
  • Developed CLI tools to streamline the configuration, fine-tuning, execution, deployment, and launching of LLM applications, speeding up the development time by 40%.
  • Designed and implemented an auto-evaluation system that understands user-profiles and personalizes prompts used to generate auto-evaluators for LLM models.
  • Created AI-assisted red teaming that generates synthetic adversarial user speeches for a specific taxonomy
  • Fine-tuned the Llama Guard model with synthetic data and got 0.1 improvement in AUPRC on the OpenAI Moderation dataset and 0.16 improvement in AUPRC on the ToxicChat.
 
 
 
 
 

Ph.D. Research Software Engineer Intern

Uber, Marketplace Pricing and Incentives

Jun 2023 – Sep 2023 California
  • Engineered an end-to-end data pipeline utilizing Spark, streamlining the process of data acquisition, cleansing, machine learning model training, and real-time predictive deployment.
  • Developed and implemented multiple Graph Neural Network (GNN) algorithms to generate unsupervised user embeddings, offering nuanced insights into complex user behaviors.
  • Architected a comprehensive graph infrastructure that utilized account features such as email and order history, facilitating an advanced analysis of interconnections among a massive 60-million user base. This contributed to a deeper understanding of user behavior patterns.
 
 
 
 
 

Computer Science Instructor (Part-time)

Sinica Education

Dec 2022 – Present Los Angels, California
  • Conducted one-on-one tutoring sessions with college-level students to provide comprehensive instruction on algorithm design, machine learning, C/C++, Java programming, and operating systems, etc.
  • Assisted students in understanding complex computer science concepts and practical application through personalized teaching methods.
  • Developed tailored lesson plans and educational materials to cater to individual learning styles and academic goals.
  • Provided guidance and support in problem-solving, code debugging, and project development to enhance students’ technical skills.
  • Monitored progress and implemented effective assessment strategies to evaluate students’ understanding and adjust instruction accordingly.
  • Fostered a positive and motivating learning environment to encourage student engagement and success in computer science coursework.
 
 
 
 
 

Machine Learning Engineer Intern/Lab Research Intern

Pinterest, Search Features

Jun 2022 – Sep 2022 Palo Alto, California
  • Successfully implemented structured search guides tailored for fashion verticals, resulting in a significant increase of 0.897% in pin clickthrough volume and a 0.195% boost in propensity.
  • Proposed and implemented advanced content-based algorithms that leveraged users’ behaviors and interests to personalize the ranking of search guides. These algorithms enhanced the user experience by delivering more relevant and engaging content.
  • Designed and developed a Deep Neural Network (DNN) model specifically for personalized query recommendations. This model effectively captured intricate patterns in user data and provided accurate and personalized recommendations for search queries, improving search relevance and user satisfaction.
  • Collaborated closely with cross-functional teams to gather requirements, conduct experiments, and iterate on models and algorithms to optimize performance and ensure scalability.
  • Conducted rigorous data analysis and experimentation to evaluate the effectiveness of implemented algorithms and models, providing actionable insights for continuous improvement and refinement.
  • Actively contributed to code reviews, ensuring high code quality, maintainability, and adherence to best practices.
 
 
 
 
 

Software Engineer Intern(Ph.D.)

Google Cloud Platform, BigQuery

Jun 2021 – Sep 2021 Remote in California
  • Designed and implemented the chatbot builder, a groundbreaking solution that generates custom chatbots capable of translating analytical English sentences to corresponding SQL queries for a given relational dataset. Achieved an impressive coverage of over 80% of existing questions and maintained an exceptional accuracy rate of over 90% among the supported questions.
  • Defined extensible intent separation criteria, employing a sophisticated approach to bootstrap intents for existing capabilities. This methodology enabled the extraction of valuable insights for future experiments and algorithmic improvements, driving the evolution of the chatbot system.
  • Implemented an advanced algorithm responsible for generating analytical training phrases in the specific format required by DialogFlow. This algorithm ensured the chatbot’s conversational capabilities by producing high-quality training data that resulted in accurate and contextually relevant responses.
  • Collaborated closely with a diverse team of researchers and engineers, actively participating in brainstorming sessions, research discussions, and code reviews to maintain the highest standards of quality and innovation.
  • Conducted comprehensive experiments and performance evaluations, leveraging data-driven insights to refine the chatbot builder and its underlying algorithms. Rigorous testing, benchmarking, and analysis were employed to achieve optimal results and continually improve the system.
  • Maintained a strong focus on staying up-to-date with the latest advancements in natural language processing, chatbot technologies, and machine learning, incorporating relevant research findings and industry best practices into the project.
 
 
 
 
 

Algorithms Instructor (Part-time)

FLAGDream

Jul 2020 – Oct 2020 Los Angeles
  • Played a key role in the development of core curricula at FLAGDream, actively contributing to the design and refinement of educational materials, assignments, and lesson plans for algorithm courses.
  • Cultivated a vibrant and inclusive learning environment, fostering a supportive community that extended beyond the formal courses. This involved organizing group discussions, study sessions, and networking events to promote knowledge sharing and collaboration among students.
  • Provided valuable guidance and mentorship to students, assisting them in successfully completing final projects that tackled real-world business problems. These projects not only showcased their technical skills but also added meaningful entries to their professional portfolios.
  • Leveraged strong communication and interpersonal skills to effectively explain complex algorithmic concepts, ensuring students comprehended and applied theoretical knowledge to practical problem-solving scenarios.
  • Demonstrated patience and adaptability in tailoring teaching methods to accommodate diverse learning styles, allowing students to grasp challenging algorithms more easily.
  • Collaborated closely with fellow instructors and staff members, actively participating in team meetings and contributing ideas for continuous improvement of the curriculum and overall student experience.
  • Stayed updated on the latest advancements in the field of algorithms and integrated relevant industry trends and practices into course materials, ensuring students received a comprehensive and up-to-date education.
 
 
 
 
 

Software Engineer Intern

Amazon Web Services (AWS), Containers

Jun 2020 – Sep 2020 Jersy City, Greater New York Area
  • Revamped Uluru (CloudFormation Registry) handlers, meticulously optimizing code to ensure seamless integration of both public and private resources while adhering to a robust object model design.
  • Took the initiative to implement and deploy the highly anticipated EnvironmentFile feature for the Task Definition resource, enhancing the versatility and functionality of the system.
  • Acted as the primary point of contact and lead in knowledge transfer, effectively transferring comprehensive knowledge of Uluru to the developer experiences team. This facilitated a seamless transition of ownership, ensuring uninterrupted development and maintenance of the system.
 
 
 
 
 

Teaching Assistant

University of California, Los Angeles

Apr 2020 – Jun 2020 Los Angeles, California
  • Assisted in developing course curriculum at CIVIL M20: Introduction to Computer Programming with MATLAB
  • Held weekly 2-hour lectures to assist students through development of homework assginments and projects that allow students to solve real scientific problems
  • Dedicated 4-5 hours per week to answer questions and facilitate a supportive and energetic community that lasts beyond the remote courses
 
 
 
 
 

Machine Learning Research Assistant

University of California, Los Angeles

Apr 2018 – Jun 2019 California
  • Built regression models (gradient descent boosting tree, artificial neural network) to extrapolate physical properties of concrete based on the empirical data (~ 10,000 observations from industry) and reduced the absolute mean percentage error of the prediction to ~9% which is the experimental uncertainty
  • Wrote a paper as the second author for ACI Materials Journal’s Special Edition on Computational Modeling

Recent Posts

Peridynamics Study Notes 1

Peridynamics basics

Boarding Pass to Mars!

Build a Deep Learning Workstation with Ubuntu 20.04

It has been a few months that I have not updated my web. I think it is a good time to write something just to wake up the sleeping …

C++ Study Note 1: const

1. const A variable or an object declared by keyword const is not motifiable. 2. Roles (1) Declare a constant variable const int LENGTH …

Hello World!

This is my first post.

Recent Publications

Quickly discover relevant content by filtering publications.

Stochastic Micromechanical Damage Model for Porous Materials under Uniaxial Tension

Despite the ubiquity of porous materials, their mechanical behaviors (e.g., fracture) remain only partially understood. Here, we …

EBOD: An ensemble-based outlier detection algorithm for noisy datasets

Real-world datasets often comprise outliers (e.g., due to operational error, intrinsic variability of the measurements, recording …

Using machine learning to predict concrete's strength: Learning from small datasets

Despite enormous efforts over the last decades to establish the relationship between concrete proportioning and strength, a robust …

Predicting Concrete’s Strength by Machine Learning: Balance between Accuracy and Complexity of Algorithms

The properties of concretes are controlled by the rate of reaction of their precursors, the chemical composition of the binding …

Recent Talks

Quickly discover relevant content by filtering talks.

IDRE2019

One property of porous materials is the pattern of their adsorption and desorption curves; in various contexts (e.g. carbon capture), …

Contact

  • yuhaili@g.ucla.edu
  • 3256 Boelter Hall, UCLA, Los Angeles, CA 90024
  • Knock the door by the elevator
  • Tue 10 am - 11 am