Led multi-disciplinary squad to productize and integrate subsystem that uses machine learning to recommend patient treatment options. Deliver high-quality product to 47 physicians in 4 locations. • Managed development of core engine which was port of IBM Research code from Python to Java. Improvements during port made code more maintainable and increased performance 100X.• Implemented RESTful microservice whose API exposes core engine functionality, such as cohort and treatment statistics and model training. Implemented in Java using Spring Framework and JAX-RS with OpenAPI (Swagger) annotations making code readable and easy to test and maintain.• Used Jackson API to implement subsystems to request, validate, and encode patient data using JSON configuration files for easy maintainability and extensibility to new medical conditions and institutions.• Worked with squad’s data scientist to develop Python based tooling for data testing which led to significant improvement in product quality.• Ran daily Agile stand-ups with squad and weekly meetings with research team and attended several other regular meetings with teams in and out of Watson Care Insights. Mentored 4 new developers. This collaboration and mentoring led to efficient development, predictable schedules, high quality software, and happy coworkers.• Generated Agile Epics and Stories with technical product manager. Led squad in sprint planning (sizing stories, assigning to sprint) and retros. Reviewed code, managed multiple code streams, and deployed to a private cloud each sprint. Applying Agile methodology as needed resulted in successful continuing multi-month pilot and study.

Team Lead and Individual Contributor, DistillerContributed to Distiller software, which searches medical records for indications of medical problems.• Improved Distiller NLP negation by investigating and integrating negation and assertion NLP software into Distiller’s Watson/UIMA pipeline.• Made Distiller highly customizable by implementing rules engine for generating indicators as Watson/UIMA annotator that supported arbitrarily complex JavaScript expressions by using JavaScript engine embedded in Java.

Led effort, in collaboration with IBM research, that integrated Sametime instant messenger with Watson Dialog Services NLP engine, Lucene and DB2 using Play Framework in Java, to implement digital assistant to answer new hire questions. • Used Lean Startup and Agile methodologies, working with HR team, UX team, and developers. Ran early pilot tests using Minimal Viable Products (MVPs) with HR and new hires. Methodologies and collaboration allowed product to be rapidly developed and dynamically changed based on user feedback.• Developed supervised mode so that HR could intercept and correct answers from system at runtime. Enabled early deployment of unperfected system and provided instant HR feedback to improve application’s answers.

Built automated system that answered questions from IBM salespeople, written in Java using Watson platform.• Managed team to ingest 1.6M documents into search indices and knowledge structures. Deployed 10-node Hadoop cluster for processing to allow timely ingestion to occur. Enhanced ingestion code in Watson Core to support new file formats and data characteristics.• Led team that improved answer precision and recall. Implemented UIMA annotators including glossary, reverse glossary, and custom ontology based lexical answer type. Implemented custom lexicon and tuned various system parameters.• Helped develop question classifier that used NLP rules and machine learning. Managed effort to create training and test data. Cowrote http://www.lrec-conf.org/proceedings/lrec2014/pdf/25_Paper.pdf and presented as poster session at LREC 2014. https://patents.google.com/patent/US9384450B1 granted.• Made Watson Sales Assistant hostable by changing Watson Sales Assistant code and generating new requirements for Watson Ecosystem hosting platform which improved Watson Ecosystem for users.

Worked on application that instantly finds IBMers using wide range of LDAP attributes. • Extended Faces Hadoop MapReduce tasks for caching IBM directory to support new LDAP attributes.• Worked with intern to integrate Faces with IBM Sametime instant messenger to provide Sametime users same great search and high-quality name completion available in Faces web UI.

Mashup Center is a web browser component aggregation system, based on the iWidget (similar to OpenSocial) specification. Written in JavaScript using the Dojo JavaScript and Dynamic HTML library; also Java running on WAS (WebSphere Application Server). Client / server communication via AJAX– specifically using JSON and APP (Atom Publishing Protocol) REST services.o Designed and implemented features of the Mashup Enabler, Builder, and Server, and iWidgets including personalization, iWidget Attributes, dynamic events, page import/export, and Atom feed parsing for the Data Viewer iWidget. o Led 8-member iWidget Team, five in China, one in Germany, and one in North Carolina. Managed and scoped requirements, reviewed all design specifications, assigned tasks, and tracked schedules. Also led the Accessibility conformance effort. Team met all schedule and functionality commitments and product met Accessibility standards without deviation.o As a Mashup Center Integration Architect, helped other IBM teams build products on the Mashup Center platform including IBM Smarter Buildings, Case Manager, and Business Space. Assisted with architecture, design, and sizing, implemented prototypes, and drove Mashup releases to meet external requirements. Also implemented prototypes for external customers, including the US Air Force.

Lotus Component Designer allows simple creation of Portlets in Eclipse using Forms, Wizards, and JavaScript that runs on the server. Runtime written in Java running on WPS (WebSphere Portal Server) and development tool written in Java as Eclipse plugins.o Designed online/offline features to support running portlets in Eclipse clients – Notes and Expeditor.o Designed and implemented data source for Lotus Connections using APP REST services.

Lotus Workplace is a set of collaboration portlets and services for WebSphere Portal Server; also an application component aggregation framework. Written in Java as JSPs and EJBs.o Designed and implemented document management portlet adapter EJB to support interfaces required by the component framework. o Led 3-member team: designed and implemented a generic desktop conversion and upload framework (patent pending) used to convert PowerPoint slides into HTML for the Workplace lightweight meetings component/portlet.o Identified, scoped, scheduled, and tracked tasks for 20 developers for the Workplace Express Product.