Improved Search Quality for non-English (NBU) languages through the use of ML translation models.

Responsible for the management, growth, and strategic direction of the company.

Responsible for the design and archictecture of the search system with an initial emphasis on content acquisition.

Designed and architected a distributed web crawler and scaled it to 1 billion documents.Frontend performance - Built a high performance message passing layer to pass messages between all frontend components (query processors, snippet servers, spelling servers, etc). Built a similated load generater that used real query logs as imput, used for doing frontend performance testing. Designed the on-disk index layout for high performance. Build high performance index reader subsystem in the query processor using Async I/O for maximum concurrency.

Third engineer hired by Cambrian to incubate what was later spun out as Kosmix.

Managed the web crawling and index group, delivered Inktomi's first 500 million document index.Designed and implemented Query execution system for the XML toolkit project. Designed and implemented a modular indexing pipeline for the web search called "superdex" to transform crawled documents into indexable data. Co-designed and implemented "slicerdicer", a tool for converting M indexes int N indexes. Implemented an index markup tool that allowed for editorial "tagging" of the index. Designed and implemented WebMap graph building software.

Worked on the VDK development team.

Worked as a software engineer in a subsidiary company, Advanced Decision Systems, that was acquired By Booz Allen & Hamilton. Developed a CORBA interface to their Minerva project, a system that aggregated and provided a unified interface to disparate data sources, including search systems and relational databases.

Member of team that developed ASTRO, a satellite resource scheduling system.