Working with Dr. Zhou Yu's lab and PhD student Ryan Shea on developing multimodal negotiation capability in AI agents as an education tool.

Monetization Team

Worked on Data Services and Telemetry team for Microsoft News (MSN), dealing

with complex data analysis on distributed data in MSN revenue pipeline, core data

pipeline and third-party partner pipeline in the petabytes range.

* Owner of revenue pipeline for all of Microsoft News (MSN) used for publisher

payout as well as A/B testing ship decision metrics.

– Ingested ~20 ad partner data on a daily basis

– Designed, implemented, and managed a separate revenue pipeline for

LeadGen revenue, including a cost-per-click model

* Owner of third-party partner pipelines which contributes to about 8% of overall

MSN daily active users

– Brought the data to a common schema with 1st party data pipelines that

empowers business reporting, content quality dashboard that measures the

quality of the recommendation system and data analyst adhoc work.

* Lead design and development of a parity tool in Spark that measures the quality of

the warm path data by comparing with the core data pipeline on cold-path.

– Defined and implemented metrics for data completeness, data quality and

latency

– Setup monitoring dashboard and alerts as well as providing troubleshooting

guide for live site mitigation

*Lead design and development of an API in Spark that allows users to consume

different datasets with different schemas and can efficiently perform filtering and

schema-alignment according to user-provided parameters.

– Increase developer agility by creating one single source for all core data

pipeline datasets.

*Drove optimizations of MSN’s core data pipeline by reducing ~ 11% of latency. The

data pipeline powers recommendation systems and model training for

personalization of MSN’s users

I worked with Dr. Olga Scrivner to write and publish a paper (as first name) titled "Persuasive Dialogue Corpus: Graph-Based Approach Combined with Persuadee Perspectives". The paper was presented at the Future Technologies Conference 2022 and won the Best Poster Award. The paper was published in the Spring Series Lecture Notes in Networks and Systems

• Worked on a dashboard that tracks user engagement metrics for local news shown by MSN and clicked on/seen by users

• Dashboad includes both user-location-based (ex. number of clicks by city that the user is from) metrics and document-based (CTR by position of document on MSN site) metrics

• Statistics taken overtime (trendlines) and per day

• Pre-processed large multiple sets of data (of size of at least 9GB each)

• Combined multiple datasets into one dataset that could be used to create visualizations using PowerBI using a pipeline

• Scheduled a pipeline to gather user engagement data from different scripts (ran once a day)

• Figured out how to get the dashboards to update automatically once a day (or even on user-demand) with data that the scheduled pipeline generated

• Presented finished project to the CVP of the Web Experiences Team, Taroon Mandhana

Impact: The finished project had a lot of visibility by the rest of the team and upper-management to help gain insight about how our local news content was doing on MSN