Guangde Chen

About Guangde Chen

LinkedIn offers me great opportunities to grow from an intern to a tech lead with management experience on multiple teams and directions. I totally believe in data and have applied many machine learning/AI technologies (Regression, GLMix, XGBoost, TF-IDF, Neural Network, transfer Learning, Constrained Optimization, BERT-based Encoder, GPT technology, etc.) to many LinkedIn core products (Home Feed, Notifications, member profiles, Guests and SEO, Content, etc.). I worked closely to a member\'s lifecycle from guests(SEO), new users (Guidance/LaunchPad), dormant users/Monthly users (Notifications/emails), churn users (Retention), and engaged users (Home feed/ Profile). I built collaborative filtering platform (Powers \"People also viewed\", etc.) and internal promo platform (To display new service or new features to members. Great for marketing teams) from end to end.Here is a short summary of my experience.1. Experienced tech lead in machine learning, AI, and statistics, with a strong track record in industry and academia, including publications in WSDM and KDD conferences and multiple granted patents.2. Passionate about integrating advanced AI and traditional AI into products. Committed to staying at the forefront of advanced AI technologies, including LLM, GenAI, Computer Vision, Video Understanding, and Multi-Modality.3. Recognized for driving business impact at LinkedIn by applying AI and utilizing data insights for major products like Home Feed, Profile, Content, Notifications/Email, Guest/SEO, Sponsored Updates.4. Strategic thinker and risk-taker, identified and acted on opportunities that moved core metrics for LinkedIn\'s Member Profile and SEO products, and played a key role in formulating solutions for Notifications, Member growth, Member retention, and Member Acquisition

Experience

1. Staff Applied Scientist and Machine Learning Engineer

   Linkedin

   May 2018 — Present

   1.Led AI initiatives for Member Profiles, overseeing strategy and team development. a) Developed and leveraged diverse member embeddings. We employed Transformer-based encoders, BERT-based, and Video encoders, optimizing cluster distances or cross-entropy loss. It involved the adoption of a two-tower framework, implementing contrastive learning or supervised learning techniques.b) Developed a profile AI platform and a Collaborative Filtering (CF) Platform, empowering various LinkedIn products2. Led pioneering AI efforts in LinkedIn SEO and Guest products, identifying opportunities and guiding a cross-functional team across AI, DS, and Engineering to develop strategic solutions. a) Constructed a cross-linking platform to address authority issues with new links for guests and bots-Implemented transfer learning and employed Term Frequency-Inverse Document Frequency (TF-IDF) for offline ranking-Utilized various embeddings for calculating similarity in candidate expansion and leveraged Embedding-Based Retrievals for efficient online updates. b) Successfully converted guests to members using conversion models with Multiple Objective Optimization (MOO). Internal promo optimization for Marketings. I worked as one of the key AI developers to build offline models (for constrained optimizations) and improve online serving for the platform to display the right service or features to the right members at the right time. Paper was published in KDD.4. Strategy-I was an early member to the formulation of retention strategies, including identifying a member’s intent, predicting the churn possibility of a member, and predicting what intervention/incentives are needed.

Education

• University of Wisconsin-Madison

  Ph.D., Statistics

  2007 — 2012

• University of Wisconsin-Madison

  Master of Science (MS), Computer Sciences

  2008 — 2011