Yu Fu Email & Phone Number

Redmond, Washington, Us

Confidential GPU

Redmond, Washington, Us

Confidential GPU

China, China, Cn

MLE at TikTok Ads Measurement Team

Seattle, Washington, Us

• Distinguished Fellow of the Artificially Intelligence for the Public Interest (AI4PI) program• Research on domain-specific knowledge graph for the public interest. Use state-of-the-art NLP methods to build knowledge graphs and extract insights from various trusted sources, including text classification, topic modeling, text summarization, named entity extraction, entity linking, etc.• Build pipelines and applications from the research to support an intelligence platform for social innovation, which is the first in the US that tracks over 2 trillion dollars of social funding and 200 million resources organized under the United Nations Sustainable Development Goals.

Shanghai, China, Cn

Shanghai, China, Cn

• Develop a predictive simulation solution that helps US healthcare insurance payers to strategize their investments for both health and economic outcomes. Core contributions including (1) Design and improve a linear optimization algorithm, which simulates the virtual populations for all US zip codes and states that correctly represent regional statistics and demographics. More than 99% of the geo-locations have an error rate within 5% for all the demographic constraints.(2) Implement and improve data mashing models and predictive models with various machine learning algorithms. (3) Migrate the project as a service to the Google Cloud Platform, which enables large-scale data generation and can respond quickly and robustly to data requests from the business team.(4) Conduct data quality assurance with model evaluation, business logic consistency check, and correlation check, ensuring that the data and following prediction models are accurate and trustworthy for clients.• Design core NLP algorithms for an automated financial document reviewer, which can correctly detect 99% of unmatched or miscalculated numbers in financial reports. The work involves table extraction from PDF document, entity extraction, context keyword extraction, word2vec retraining, and RNN model design.• Lead the standardization and automation transformation of the company's computer vision team. Build a Python library covering the most common functions used in computer vision projects, which simplifies the data preparation, model training, and model evaluation, leading to faster project iterations and delivery.• Provide support for the auto invoice reimbursement platform of travel expenses. Major work involves code and algorithm reviews, as well as the evaluation and selection of deep learning network models.

Beijing, Beijing, Cn

• Apply deep learning models (SSD, Faster-RCNN, U-Net) to classify, detect, and segment illness displayed on medical images.• Build a non-medical image filter through a modification of image search. The filter uses HSV histogram and LBP and HOG texture as features, and the searching algorithm is implemented using LSH (Local Sensitive Hashing). Achieve <1s response time and both precision and recall at 0.97.• Build a quality control system that gives scores of darkness, brightness, edge light leakage, and medical images' clearness.

• Apply computer vision and machine learning techniques to build and improve algorithms of face recognition and facial feature points alignment.• Design and build facial feature points labeling tool. The tool gives labeling suggestions predicted by the existing model. Also build a quality control tool to check labeling quality.• Reduce training time to half by designing a grid hyper-parameter binary search algorithm.

Master Of Science - Ms, Computer Science: Data Science Specialization

Bachelor’S Degree, Applied And Computing Mathematical Science: Discrete Math And Algorithms