Stanford University

Duke University

Research And Teaching Assistant at Stanford University

Stealth Startup

AI Research Engineer

- Developed an innovative solution using fine-tuned LLMs via both instruct fine tuning and RLHF to efficiently extract product information from unstructured data.
- Introduced reinforcement learning from weak human feedback (RLWHF) approach to reduce the cost
of collecting human preference data.
- Pioneered deployment strategies for fine-tuned Large Language Models (LLMs) using Kubeflow and Google Vertex pipeline.

Wayfair

Machine Learning Scientist

- Led a team of five PhDs and one JD, achieving the first-place in the Wayfair innovation sprint.
- Developed an information extraction pipeline that connects images with domain-specific text using CLIP. Achieved a 71.4% improvement over the specific trained ResNet models, optimizing product identification for the website.
- Prototyped foundation models for cost-effective and efficient model training, providing a scalable solution.

Data Science & Machine Learning

- Developed a robust, scalable system for engineers to uncover crucial EV-battery deposits from aeromagnetic data, leveraging deep learning models. This resulted in an eightfold enhancement in the sharpness of original aeromagnetic images.
- Designed a framework using adapted GAN models to generate realistic, high-resolution magnetic images, with a focus on preserving large spatial features.

KoBold Metals

Data Scientist

Research And Teaching Assistant

Graduate Teaching Assistant

yizheng wang