My Work in Data, Machine Learning & AI

My client, a major American hotel chain, wanted to develop and implement demand-based pricing models to optimize revenue. My team was engaged to create an implementable concept that was validated by internal users and ready for development.

I was responsible for understanding and capturing both user (hotel owner, corporate owners) and business requirements and ensuring viability with the data scientists and business owner; this required significant research effort and cross-functional team collaboration.

Specifically, I created and validated the user experience through prototypes using data outputs from data team to illustrate impacts of demand-based pricing models for the business owner. These pricing models were eventually put into production, with an estimated 7-figure revenue increase.

My employer wanted to explore innovative ways to engage with machine learning. As a leader of the organization's innovation team, I led a group of interns from Vanderbilt University's Innovation Lab to conduct research, create the design, and build an initial proof of concept of a machine learning-driven Employee Experience platform.

I provided hands-on guidance for identifying the business problem and related strategy and communication materials outlining the expected business benefit for both employees and people managers. I was a member of the technical team, partnering with team members from both Vanderbilt University and Microsoft to develop the proof of concept.

The POC successfully and accurately leveraged random forest machine learning model to provide predictions of those employees at high risk for leaving the organization based on actual data, as well as recommended retention strategies/approaches for managers of these employees.

As a leader of both the internal innovation team and owner of the enterprise knowledge management system, I wanted to test the viability of a knowledge graph-driven content recommendation engine.

Using an existing knowledge management system, my developer and I worked to create a proof of concept content recommendation system powered by a knowledge graph. This accomplished a task in moments when previously it took a single person multiple hours per week to track and manually link the content.

This knowledge graph utilized the existing taxonomy connected to an ontology that we had created to identify key connections between concepts. A graph database served to house the connections. When completed, the proof of concept successfully pulled related content from disparate areas of the site and displayed it on the user's original screen.

We presented this process to the enterprise with recommendations for applying it to their daily work. This work was shared with the enterprise development team and was entered into the product roadmap for implementation.

My customer, a global energy producer, wanted to explore ways to offset carbon emissions in pursuit of their enterprise climate goals.

I served as project lead and client liaison for this effort that developed a proof of concept that leveraged AI to identify and remove duplicate content from their enterprise knowledge management content. It was hypothesized that a reduction in content would reduce server load, thus reducing carbon emissions needed to support the content.

My team constructed a vector database that converted content into integers. An AI algorithm then evaluated those integers for duplicates. The system successfully identified duplicate content and our estimates showed that even with a reasonably small POC sample size, the net carbon reduction was greater than the carbon footprint to complete the analysis.

When conservatively applied to the client's very large content collection, the potential carbon reduction was equal to 20 flights between JFK and Heathrow airports.

I developed the strategy for implementing this technology into the enterprise environment. This strategy included near and longer-term implementation roadmaps and prioritized automation features that could proactively identify and prevent duplicate content, thereby reducing the level of effort on the users of the system.

My client, a large salary data clearinghouse, sought to enable self-service data access and reporting for their customers.

I worked with users to identify and understand user needs regarding salary data access and reporting. I worked with internal organizational stakeholders to understand current process flows and bottlenecks.

I used this information to develop a functional prototype proof of concept for the future state self-service data reporting tool, which I presented to both internal stakeholders and users for validation. Following successful validation, I documented the process changes and requirements necessary to enable this shift from high-touch manual effort to self-service digital effort.

This effort included a not jus the requirements and visual elements necessary for implementation by the client's technical team, but also a full implementation roadmap and comprehensive change management and training plans to ensure that employees who had been completing report creation as a service role-based activity could successfully pivot to customer support on a self-service platform.