Ana Mamatelashvili, PhD

Our team designed, implemented, and deployed to production an information extraction and post-processing solution to automate insurance claims processing.

Our application receives an image, sends it through an OCR component, processes the raw OCR output according to the insurance company's business logic, and prepopulates the claim form with all the fields—like price, service item, provider, and more.

The team consisted of a machine learning engineer—for deployment and integration—and myself as a data scientist— performing BI research, problem scoping, and solution design and implementation.

The solution achieved a 30% reduction in average handling time for human review of the claims.

I conducted a technology review, project scoping, and use case development for a chatbot for one of the big four Australian banks.

I worked across several teams to achieve this, including security, cloud engineering, and customer service teams.

I coached a data scientist to build a general solution for utilizing rich insurance member and claims data to support digital sales and healthier lifestyle campaigns.

The project was deployed on DataBricks with R.

I designed and implemented a deployment-ready custom model iteration and evaluation process for a chatbot for a major insurance company. The new process was integrated a couple of months later.

I worked closely with the chatbot maintenance team to identify the best suitable solution.

This project focused on studying the online presence of the Georgian labor market. The project's scope extended to identifying the largest and the most reliable data sources, developing methodologies and implementing algorithms to collect, consolidate, and maintain the data, designing mathematical and statistical methods to analyze the data, and publishing results and conclusions on the project's website.

In this project, I focused on studying strong arcwise connectedness properties in nite graphs and related spaces. Among other work, I established a threshold connectedness degree of seven points and developed methods and algorithms to characterize graphs of lower connectedness degrees.

Publications (besides the one available in the link above):
• https://arxiv.org/pdf/1209.5454.pdf
• https://arxiv.org/abs/1801.00179

This project was dedicated to studying order structures arising from real numbers and related spaces. Among other work, I developed an algorithm to recursively construct the maximal possible sized collection of ordered structures from the real numbers with differing order complexities, consequently answering a 40-year-old question in foundations of mathematics.

Publications (besides the one available in the link above):
• https://link.springer.com/article/10.1007/s11083-017-9423-6?wt_mc=Internal.Event.1.SEM.ArticleAuthorOnlineFirst
• https://arxiv.org/pdf/1606.06493.pdf