Goals

  • I build high-performing AI systems to solve real-world challenges.

  • Build a data engine to distill human expertises and knowledge to build the state of the art AI systems.

  • Be a life long learner.

Work Experience

2023 - 2024 | Senior Machine Learning Engineer, Uber

  • Leads the development of Uber’s Multi-modal Foundation Model for document automation. I oversee all stages of development, including data curation, model design, training, evaluation, fine-tuning pipeline, and serving. Here is a presentation I gave at Ray Summit 2024.

  • Led the development of a few pilot consumer-facing Generative AI applications in the company, including the new shopping chatbot for UberEats and copilot AI Assistant for Uber drivers.

  • Led the design and development of a robust, scalable distributed training framework for pretraining and fine-tuning Large Language Models, with PyTorch, Ray, DeepSpeed, and Kubernetes.

2023 - 2024 | Part-time Research Engineer, Stanford CRFM

  • Training and evaluate large language models with TPU and Jax.

  • Contribute to Levanter, a Jax-based codebase for training foundation models.

  • Contributor to RedPajama-1T dataset and models.

2018 - 2023 | Senior Machine Learning Engineer, Landing AI

  • Developing LandingLens, an end-to-end MLOps platform for computer vision applications. I’ve led the work on implementing object detection and semantic segmentation models, developing AutoML techniques, managing edge device deployment, and improving error analysis at evaluation.

  • Working directly with more than 10 enterprise customers in North America, Europe, and Asia to help them build successful ML applications. I develop playbooks and documentations to get customers set up for long-term success.

  • Provide technical mentorship for engineers on ML, computer vision, cameras, and software development.

  • Working directly with Andrew Ng to develop and advocate the Data-Centric ML methodology to the global ML community.

2018 - 2018 | Machine Learning Engineer, Shopify

  • Image Search and Personalization in production

  • End-to-End data science support for a fast-growing photo marketplace (Shopify's Burst), from setting up data infrastructure to algorithm implementation and product analytics

  • Machine Learning research: publication in ACM RecSys 2018

2016 - 2018 | Machine Learning Researcher, University of Toronto

  • Published a paper in International Joint Conference of Artificial Intelligence (IJCAI 2017) on non-linear planning with Deep Net Transition Models as a co-author.

  • Published an Honored Thesis on the Recommender System.

  • Implemented Deep Reinforcement Learning in large scale system planning.

  • Implemented and evaluated the performance of anomaly detectors with Deep Autoencoder, PCA, and RPCA.

  • Supervised by Professor Scott Sanner in the Data-Driven Decision Making Lab.

2017 | Research projects at Mozilla Firefox and 500px

  • Develop machine learning systems to recommend 500px users with photos that they will most likely to enjoy with.

  • Developed an end-to-end user feedback analytics engine for Mozilla Firefox team to process tens of thousands of user feedbacks in real time, with multiple components including translation, sentiment analysis, topic categorization, and dashboard visualization.

Education

MS Computer Science from Stanford

Focus on AI: Computer vision, generative models, robotics, and natural language processing.

BASc Industrial Engineering from University of Toronto

Focus on Operations Research and Machine Learning.
Publication on IJCAI and ACM RecSys.

Public Talks

Contact Me