Summary

Goal: lead position with opportunities for conception, research, design and implementation of groundbreaking web services and systems. Have built large, complex distributed, multimedia and web-related systems. A particular forte is quickly grasping and debugging complicated third party product source. Enjoy working with folks of diverse backgrounds. Constantly striving to innovate and learn and interested in organizations that encourage both.

I've worked in a variety of areas including web and enterprise search, information retrieval, data mining, video processing, video on demand, distributed systems, content classification and management, UI design, system administration and database systems. Please e-mail if you'd like a complete CV.

Quick Overview of Experience

Lead developer/architect for contextual ad targeting for web pages project at Looksmart. Tests show our approach (jointly developed with a colleague) to hold great promise versus Google's AdSense. Single-handedly conceptualized, architected and authored backend system using Lucene, Java, MySQL, Linux, Tomcat and various public libraries. System performs ad/url management as well as matching and serving of ads in response to webpage accesses. Currently moving system to beta with product rollout at the end of the year.

Prior to this, I was responsible for the link score component of the WiseNut.com web search engine ranking algorithm. This involved collecting and mining large amounts of data from a variety of sources, producing and testing new approaches and tools to tune existing ones. The changes incorporated into WiseNut helped maintain its relevance vis a vis other search engines.

Technical Group Lead for Categorization in Enterprise Search (UltraSeek.com):

Engineering lead during Quiver Classifier evaluation process (built using 100K+ lines of Java, MATLAB, J2EE Servlets, JSP, JDBC, XML, C++, and Oracle/MSSQL). After acquisition (Aug 2002) and unavailability of all 8 Quiver engineers, recruited team members, performed performance analyses and fixed critical bugs. Team achieved first product release in Nov 2002. Also developed a SOAP based web service extension (via Python, C) to Ultraseek and rewrote significant portions of the UltraSeek Content Classification Engine in C. Added display UI features.

Web Search Division:

Developed a system, GigaDoc, using C++ and Tcl, as part of a team of three, to crawl entire web and create billion document index. Created an internal portal, (HTML, CGI, Perl, Oracle) for storing, browsing, and bulk-extracting porn judgments on various web sites. Worked on Media Search, a system to scour the web for MP3 files.

Designed and implemented in C the first publicly available fast compressed domain MPEG to JPEG transcoder. Software outperforms spatial domain counterpart by factor of 2-3.

Published the first survey of video data on the web. Developed Tcl and C based distributed automated system to locate, download, and analyze this data (about 100 GB worth of QuickTime, MPEG, and AVI files).

Analyzed server log files, via Tcl scripts and Excel, of an ongoing Video on the Web experiment to determine how users access videos. First web related study of its type.

Created the first distributed web caching system to focus exclusively on video. Used Jigsaw (Java based web server), Java, and Tcl to simulate and prototype project.

Contributor to Continuous Media Toolkit (CMT), a Tcl/C toolkit for creating distributed multimedia applications.

Maintained and developed, as member of a team, an object oriented InterViews/X11 based C++ toolkit for creating user interfaces and graphical applications. Also, extended Glish, a C++ based system for distributed accelerator control sequencing, to control remote clients.

Created C like language, interpreter and library package to implement an user-interface to a particle accelerator data management system. Ported package to multiple architectures including SYBASE.

As member of a team, developed a suite of C test routines and prototype applications in order to evaluate a variety of real-time kernels including VxWorks and Lynx.

Responsible, as part of a group, for developing and maintaining a data acquisition system for a non-linear dynamics Physics experiment at FermiLab.

Education

Thesis Advisor: Brian Smith (CS Dept)
Title: Techniques for Improving Multimedia Communication over Wide Area Networks