Beijing, China
Highlights --(1)Creative. (2) Detail-oriented. (3) Critical Thinking. (4) Quick Learning and Self-reflection. (5) Sensitive to data. -- A deep passion in carrying out research on algorithms and improving them. Hope to become an excellent data scientist who makes the algorithms better.
--Tracked cutting-edge research including knowledge-enhanced pretraining、prompt-learning and other Natural Language Understanding technologies --Built models to solve problems of advertisements badcase.
Led a small team and focused on algorithms optimization and cutting-edge research.
EWELL is a publicly listed tech company for medical industry. I work in AI department of the company R & D team. My responsibilities: (1) Iterated and improved business models. (2) Initiate research projects
Knowlegene is a financial tech company. I work in AI Lab department, explore current AI technologies and initiate research projects.I also work with coworkers of business and development department for business products development. --Researched on Generative Adversarial Network and automatic summarization by exploring papers and tracking advanced AI technologies. -- Improved current Generative Adversarial Network to generate better texts and Chinese financial news. --Created a state-of-the-artGAN model based on attention mechanism with TensorFlow, which applies reinforcement learning to generate texts ,decreasing convergence epochs and increasing quality of generated texts by more than10% over comparable state-of-the-art models. -- Wrote a paper related to above model and submitted to the top AI conference AAAI (7,6,4 score). --Build a prototype extracting automatic summarization for Chinese financial news. (1) Enhanced segmentation methods of Chinese words. (2) Strengthened performance of word embedding.(3) Refined similarity measures between sentences. Allowing 14 of 16 data were overlapped between the algorithm and human labels. -- Presented findings and delivered a public lecture to skateholders; Discussed actively in paper workshops for different AI topics
--Utilized mean.js (i.e. Node.js, Angular.js, MongoDB, and Express) to develop in collaboration with other 3 members for an app, which applied machine learning to predict patient’s diabetes and now has over 50 users. --Customized User interfaces such as blog interface where users are able to comment and vote for blogs. --Improved feature selection methods for food classification, eliminating redundancies among food elements and increasing classification accuracy by 2.7% .
--Classified subtypes of lung cancer survival data with R, which has high dimension gene feature spaces, survival time and censoring statues indicating if the survival time is completed. --Optimized feature selection to discover optimum gene set containing 20 genes by using unsupervised clustering and statistical modeling, which matched well with literature description . --Built and maximized performance of models based on semi-supervised and unsupervised to classify subtypes of lung cancer data. --Created a new version of the SPC (i.e. supervised principal components) approach, which obtained the best classification performance with a p value of 1.35e-06 comparing to investigated state-of-the-art methods.