• Build deep learning models (LSTM, CNN, GAN) for anomaly detection on LinkedIn business and engineering metrics, monitored on the ThirdEye platform• Formulate innovation in existing statistical algorithms to solve pain points in using ThirdEye• Estimate system traffic headroom in a data-driven way to optimize resource allocation

- Developed Photo Grader, a tool to detect photo quality, for an online photographer booking platform- Trained and fine-tuned a MobileNet based Convolutional Neural Net model on a labeled aesthetic database using GPU; built a docker image for deployment- Measured the business impact of increasing photo quality and improving other engineered features; provided strategic advice on streamlining the platform- Demo: photograder.wangruoying.com, Blog: goo.gl/pZucxb

- Developed Python routines to conduct fuzzy matching of firms over time and across datasets- Studied the causal relationship between import competition and innovation behavior of ~800,000 firms- Feature engineered instrument variables for foreign competition and designed falsification tests- Developed a theory on import competition and innovation to guide welfare analysis- Led tutorial sessions and designed course materials tailored to undergraduate students with a range of academic background

- Worked with confidential firm data to study the productivity of ~100,000 Canadian importers- Developed a control function methodology to identify how buyer-supplier relationship could affect firm performance