Multimedia Relevance (Team Lead):Established and led the team responsible for the ranking and relevance of Bing image and video search. Team lead in charge of planning, overseeing and assisting in projects, mentoring team members, holding team meetings, hiring, and conducting performance reviews. Proposed, architected, and oversaw numerous projects including, regular releases of machine learned rankers using neural networks and decision trees, satisfaction models that predict relevance of individual search results based on ML models of user interaction, preference of popular results using click prediction, use of image understanding signals, near duplicate removal, ranking and query classification for fresh/news queries, performance improvement via multi-stage ranking. Oversaw judge based relevance measurement. Developed quality metrics that utilize both relevance and secondary factors such as image appeal and video popularity. Regularly engaged in cross-team collaboration including, STC-Asia, STC-Europe, MSR, and other product groups.Multimedia Relevance (Individual Contributor):Sole person responsible for Bing Image and Video ranking and relevance. Participated in the first release of Microsoft’s image and video search (Live Search). Trained the first non-research image ranking model and first ever video ranking model using Lambda Rank neural networks. Architected and collaborated to create the features used for multimedia ranking. These include text matching features (BM25, page understanding, page layout, word proximity, etc.), user event “click” features, image content features, and a variety of other signals. Responsible for providing regular improvements and ranker releases. Developed class specific ranking for a number of high value query segments, including both query classifiers and custom ranking. Oversaw judge based relevance labeling, defined guidelines, maintained infrastructure, built training sets and performed competitive measurements.

Flat-Panel 3D Imaging (Cone Beam CT / Tomosynthesis):Instigated an exploratory project in this area. Focus on 3D reconstruction, prototyping, and new product development.Computer Aided Detection:Team leader in charge of new research on automated cancer detection for mammography. Developed an image registration algorithm for bilateral and temporal alignment and differencing. Also developed a lung segmentation routine for lung CAD, and a tooth segmentation routine for dental shade matching.Facial Feature Finding:Developed a texture coherent active shape model algorithm for finding objects/features of interest in an image. The primary focus of this work was to develop a mechanism for detecting facial features in portrait images. The algorithm has been used in both products and research for automatic portrait retouching, red-eye detection, eye tracking in an immersive mid-air refueling simulator, skin segmentation, facial animation, and gender/age classification.Facial Imaging:Worked with a product concept team on facial animation. Implemented algorithms for fast face detection, gender categorization based on principal components analysis, and probabilistic skin segmentation.Image Morphing:Implemented routines for image morphing and warping, fundamental matrix estimation, and view-morphing.

Teaching and Research associate.

Windows Printer Driver Development:Developed a Windows printer driver to interface with Xerox’s Common Imaging Interface for 2D rendering.

Robotic Library Vision System:Developed a neural network based system for adaptively controlling the brightness and contrast of the vision subsystem of a robotic tape library.